Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Research ArticleAn HTML5-Based Pure Website Solution for Rapidly Viewingand Processing Large-Scale 3D Medical Volume Reconstructionon Mobile Internet
Liang Qiao12 Xin Chen3 Ye Zhang1 Jingna Zhang1 Yi Wu4 Ying Li4 Xuemei Mo4
Wei Chen5 Bing Xie5 andMingguo Qiu1
1Department of Medical Image College of Biomedical Engineering Third Military Medical University Chongqing China2Department of Computer Science College of Biomedical Engineering Third Military Medical University Chongqing China3College of Software and Computer Chongqing Institute of Engineering Chongqing China4Department of Digital Medicine College of Biomedical Engineering Third Military Medical University Chongqing China5Department of Radiology Southwest Hospital Third Military Medical University Chongqing China
Correspondence should be addressed to Mingguo Qiu bobqiao163com
Received 10 January 2017 Revised 24 March 2017 Accepted 13 April 2017 Published 30 May 2017
Academic Editor Manolis Tsiknakis
Copyright copy 2017 Liang Qiao et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
This study aimed to propose a pure web-based solution to serve users to access large-scale 3D medical volume anywhere withgood user experience and complete details A novel solution of theMaster-Slave interaction mode was proposed which absorbedadvantages of remote volume rendering and surface rendering On server side we designed a message-responding mechanism tolisten to interactive requests from clients (Slavemodel) and to guideMaster volume rendering On client side we used HTML5 tonormalize user-interactive behaviors on Slave model and enhance the accuracy of behavior request and user-friendly experienceThe results showed that more than four independent tasks (each with a data size of 2494MB) could be simultaneously carriedout with a 100-KBps client bandwidth (extreme test) the first loading time was lt12 s and the response time of each behaviorrequest for final high quality image remained at approximately 1 s while the peak value of bandwidth was lt50-KBps Meanwhilethe FPS value for each client was ge40This solution could serve the users by rapidly accessing the application via oneURL hyperlinkwithout special software and hardware requirement in a diversified network environment and could be easily integrated into othertelemedical systems seamlessly
1 Introduction
The 3D visualized reconstruction which is built from sec-tionalmedical images (SMIs) such as CT has an advantage ofdirectly displaying the focus of a disease and is widely used inmany fields including diagnoses case discussion educationand patient consultation [1 2] In the era of mobile Internet(WIFI or 3G as fundamental carrier) with increasinglylarge amounts of SMI data of high quality people desire torapidly view and process this 3D visualized reconstructionof high quality in a diversified network environment viadifferent client devices without installing special software [3]In particular many studies have reported that there wereno significant differences of diagnostic accuracy regardingmore and more diseases between mobile devices and the
traditional workstations that promoted the further develop-ment of medical imaging technology on mobile Internet [4ndash9] Presently the 3D visualized reconstruction techniques inmedicine have fallen into the following two main categoriessurface rendering and volume rendering [10] both with theirown characteristics to adapt to the Internet application
Surface rendering in medicine extracts the interestedsurface contour of the anatomical representation from SMIsinto a 3D geometrical model which is usually stored as aVRML VTK or X3D file [1 11 12] Some Web3D technolo-gies such as WebGL could directly support these modelsvia most of the major web browsers [13] with a very goodnative user experience and that is a popular way to utilizeInternet applications for medical 3D visualization Howeverbecause the information integrity depends on presetting the
HindawiInternational Journal of Telemedicine and ApplicationsVolume 2017 Article ID 4074137 13 pageshttpsdoiorg10115520174074137
2 International Journal of Telemedicine and Applications
interested area the client user cannot arbitrarily view theinternal anatomy structure or analyze the complex orga-nization relationship Therefore surface rendering technol-ogy has been mainly used for special disease with specificrequirements [11 12 14] and it could not adapt to thecustom behavior of the client user in mobile Internet verywell
Volume rendering describes a wide range of techniques togenerate images from 3D scalar data [15] It is very convenientto view the internal anatomy structure and describe thecomplex organizational relationship through user adjustmentof the parameters of color and opacity on each voxel [16]However volume rendering requires a complete computationof the original volumetric data for each userrsquos behavior eventthe cost of computation being much higher than that for sur-face rendering Thus volume rendering technology is oftendeployed on a dedicated image workstation and in a privatenetwork environment where the scope of application is lim-ited Some low-level graphics APIs such as OpenGL ES 20have been integrated into WebGL for Internet applicationwhich could be theoretically enabled to support GPU-baseddirect volume ray-casting implementations or support someslice-based sampling techniques to get close to the imagingresult of ray-casting-based implementations Neverthelessthe imaging quality and efficiency is very limited not justbecause of performance of common GPU but also becauseof the limited performance of web pages [7ndash9] Further-more there is still high requirement to network for originalvolumetric data transmission Currently there is a remotevolume rendering scheme to solve the limit This schemedeploys the volume rendering job on the server side to a real-time response to behavior instructions from the client andtransmits the 2D projection image from volume renderingto the client display where the client is only an interface forsending instructions and displaying results [9 17ndash20] Forexample the user slides the mouse from point A to point B toexpress an operation of 3D rotation on the client side and theserver just calculates the final viewing angle from A to B andsends the final result to the user Because there is no originaldata downloading and no rendering cost on the client sidethe network load and computing pressure of the client couldbe significantly reduced However this interaction model ofldquothe client sends one operation request the server replies oneresult at a timerdquomay cause a bad user experience Specificallyit may lose the value of Frames Per Second (FPS) one of themost important parameters of the 3D interactive experienceTherefore all of [9 17ndash20] discretized the trajectory of onesingle operation request from the client into several requestpoints to request continuous responses from the server Forexample if the path between mouse points A and B wereconverted into several separated remote volume renderingrequests from the client the sever would respond to eachrequest and send continuous 2D projection images to theclient Thus the value of FPS could be increased howeverit may lead to a huge pressure on the server load (responseto a large amount of requests) and network load (continuousdata stream) Additionally [9 17ndash19] ask the client to installcustomized software or plug-ins which may lead to a lack ofcompatibility Therefore existing remote volume rendering
methods do not adapt to themobile Internet application char-acteristics of lightweight [21] very well which ask for lowerresources of network and computing and avoid unnecessaryinstallation
In this study we plan to combine the advantage of surfacerendering technology based on WebGL with the advantageof remote volume rendering to design a solution for viewingand processing 3D medical volume reconstruction in a purewebpage environment Finally a visualization interactionplatform would be built This platform could be used any-where in a diversified network environment and diversifiedclient devices with a good user experience similar to thenative application
2 Computational Methods and Theory
The core concept includes the Slave model and Mastervolume The Slave model is a contour model rendered bysurface rendering technology running on WebGL as a 3Dinteractive interface on the client side The Master volumeis a volumetric data rendering from original SMIs on theserver side The final required image of high quality is aprojection drawing from theMaster volume according to thebehavior instructions from the Slave model This solutioncould be named the Master-Slave dual-channel interactionmode and is expected to enhance the user experience similarto the native application in a pure web page to reduce thepressure of the server load and network load and to meetthe requirement of lightweight volumetric data sharing in amobile Internet application
21 Design of the Master-Slave Dual-Channel InteractionMode The architecture of this solution constitutes the fol-lowing two parts the server side and client side The serverside renders theMaster volume to generate the final requiredimage of high quality while the client side provides the Slavemodelrsquos interactive navigation
The detailed procedures are as follows (schematic dia-gram is shown in Figure 1)
On the server side in zone A groups of SMIs couldbe uploaded to the server through the network interface ofPACS or removable storage medium Thereafter the relevantSlave models could be automatically pregenerated by roughcontour extraction The listener in zone B accepts theuserrsquos web behavior request (instructions) in real-time andguides zone C to load and render the relevant SMIs inmemory which is defined as the Master volume generatinga 2D projection image from the rendered result for clientdownloading
On the client side while the user wants to view andprocess a group of SMIs the client could download thecorresponding Slavemodel from the server and reconstruct itlocally in zoneD meanwhile request the server to load andrender relevant SMIs in memory Thereafter in zone D theuser could view and process the local Slavemodel arbitrarilyand submit the final behavior request (instructions) to theserver to acquire the required projection image from theMaster volume in zoneB C E
International Journal of Telemedicine and Applications 3
Interactivebehaviors
Sectional medical images (SMIs) amp Slave model
Service
N
M
Any device amp OS amp browser
e screen eect
2D imageMaster volume
Listener
Projection
Slave model
Internet
Client
Volume rendering
Behaviorrequest
Request Mgmt
Loading in memory
1
2
3
85
86
middot middot middot
middot middot middot
middot middot middot
12
3
85
86
12
3
8586
middot middot middot
①
④
⑤
③
②
Figure 1 Architecture of the pure web-based solution via theMaster-Slave dual-channel interaction mode
Here the client could be a pure web browser running ondiversified devices supporting the WebGL technique Thatmeans there is no installation for the client
22 Rendering of the Master Volume We adopted the ray-casting algorithm [22] to reconstruct the Master volume ofSMIsThe algorithm is a direct volume renderingmethod thathas been widely used for radiodiagnosis but requires a highcomputing performance of the device First we transformedthe SMIs into volume voxels in 3D space (Figure 2(a)) andthen simulated rays to cast each voxel with a customizedcolor and opacity [23] into 2D space (Figure 2(b)) Thismethod could be used to describe a complex organizationrelationship Furthermore it could display internal anatomystructure via removing (clipping) part of the voxels thatsimulate a userrsquos behavior of dissection (Figure 2(c))
Here we directly used the VTK (Visualization Toolkit)library [24] to realize the ray-casting algorithm For cus-tomized color and opacity several frequently used schemescould be deployed on the server side in advance a dropdownlist was designed on the user interface to correspond to eachscheme for remote volume rendering [25 26] For clippingVTK provides a class of vtkPlane to build a virtual Clippingplane to guide the voxel removal [27] It means the clippingjob only needs the specific location of the virtual Clippingplane in the 3D coordinate systemThe specific location couldbe described by the following two parameters in the computergraphics method normal vector from the origin point to theplane and the coordinate position of a point on the Clippingplane both from the client submission
23 Generation and Interaction of the Slave Model The Slavemodel was pregenerated on the server side and was designedto be a navigational tool to interact with theWeb page Whileclients want to view and process theMaster volume the Slavemodel could be downloaded to the client and reconstructedon the client side
231 Generation and Slimming of the SlaveModel To interactwith mobile Internet the data size of the model should beas small as possible (much slimmer) while the clarity ofthe contour is acceptable However using the traditionalcompression-decompression algorithm may ask the client toinstall additional software Hence the steps in the generationof a smaller Slavemodel can be designed as follows (Figure 3)Here we used a group of SMIs of the head and neck CT asan example and the resolution of voxels is 512 times 512 times 495249MB (testing data (1) in Section 4)
Step 1 (resampling the original SMIs for first slimming) Loadthe corresponding SMIs into memory shrink each sectionalimage to a standard size of 32 lowast 32 via the resampling of the2D space (Figures 3(a) and 3(b))
Step 2 (building the 3D geometrical contour model) Amarching cubes algorithm [28] one of the most commonmethods of surface rendering was used to extract theinterested surface contour of the anatomical representationfrom resampling SMIs into a 3D geometrical contour modelHere according to the feature of the CT grayscale [29] weset a particular contour value at minus800 or 200 that couldseparate the skin or skeleton of the body from the resampling
4 International Journal of Telemedicine and Applications
(a) Voxels in volume data (b) Pseudo color amp opacity for each voxel (c) Clipping
Figure 2 Schematic diagram of the direct volume rendering method
Size YSize X
Size ZSize Z
Size XSize Y
Y Y
X X
Z Z
(a) Original SMIs
Step 1 Resampling the original SMIs for rst compression
Shrink each slice
(b) Resampling amp strech
(c1) Contour value minus800
(c2) Contour value 200Size 526MBStep 2 Building
3D contour model
(d1) Polygons 50 Polygons 10
(d2) Polygons 50 Polygons 10
Step 3 Reduce the triangles
Step 4Saved asvtk le
(eg 512 times 512 times 495 249 MB) 311MBSize 104MBSize
333MBSize 913MBSize674MBSize
2 times 32 times 495)(eg 3
Figure 3 Schematic diagram of the generation and compression of the Slavemodel
SMIs respectively to acquire a skin model or skeleton model(Figures 3(c1) and 3(c2)) In this case the data size of the skinmodel and skeleton model could be reduced to 674MB and526MB respectively
Step 3 (reducing the total number of polygons for secondslimming) The 3D geometrical contour model from Step 2is constitutive of polygons and we could reduce the totalnumber of polygons to further reduce the data size In thiscase the number was reduced to 50 and 10 and thedata sizes were reduced to approximately 3MB and 1MBrespectively (Figures 3(d1) and 3(d2))
Step 4 (saved as vtk model file) The slimmed model wouldbe saved in a file of the vtk format which is a storage formatof the geometrical model supported by WebGL and thenstored in the same directory with original SMIs
The result of Figures 3(d1) and 3(d2) preliminarily showedthat the Slave model whose data size is approximately 1MBstill could be fit for interactive navigation (more experimentsdetailed in Section 41) Moreover an additional mutualinterface of the client was designed as a remote parametersetting of the contour value and the total number of polygonsto permit users to customize the Slavemodel as desired
232 Design and Quantization of the Interactive Behaviors inthe Slave Model For a friendly interactive experience on theclient side a simple style of interactive behaviors of the clientand corresponding quantitative methods should be designed
to guide remote volume rendering including the viewingangle request and clipping request
For the Viewing Angle Request The interaction of viewingangle includes rotation flipping zooming and panning [30]We defined the interactive 3D scene including only one Slavemodel and one camera and customized an interactive modeof the camera in VR technology to meet our requirementFigure 4(a) shows the principle of the interaction of theviewing angle between the camera and Slave model Welimited the lens of the camera to always be taken on point1198751 (0 0 0) which was the origin of the world coordinatesystem The interactive behaviors of the viewing angle werecarried out by only changing the camera position exceptpanning For panning it was carried out by changing theSlavemodel position see examples in Figure 4(b)
According to the definition while sending requestinstructions to the server for remote volume rendering theinstructions only include three quantization parameters fromthe client to accurately describe the interactive behavior ofthe viewing angle request to the Slavemodel The three quan-tization parameters include the following attribute view updirection attribute position of camera and attribute positionof the center point of the Slavemodel
Here attribute view up direction of camera indicates thedirection of the top of the camera For example (0 minus1 0)a value of the View Up direction indicates the camera isperpendicular to the XZ plane in the world coordinatesystem and the top of the camera is in the opposite direction
International Journal of Telemedicine and Applications 5
(a) Principle of interaction between camera and Slavemodel
(i) Original camera view
(iii) Zoom via shorteningthe distance from CaA to P1
(iv) Flip from CaA to CaB (v) Panning via movingthe center point of Slavemodel from P1 to P2
(ii) Rotate around axis CaA-P1 (Y-axis)
(b) Example of interactions of viewing angle request
Figure 4 Principle of the interaction of the viewing angle between the camera and Slavemodel and the corresponding example
along the 119885-axis (see the relationship between camera CaAand Slave model in Figure 4(a) and corresponding example(i) original camera view in Figure 4(b))
In fact WebGL has provided some interaction methodsof the moving camera to view 3D objects via mouse- andgesture-based interactive behaviors which were packagedinto a library of TrackballControlsjs for a web page scriptWe could directly utilize the js library to acquire the threequantization parameters However we need to rewrite thejs library to restrict and customize the interactive modeof the camera that we have defined to ensure the consis-tency of the interactive rule between the Master and Slaveside
For the Clipping Request For arbitrarily clipping and viewingthe internal anatomy structure of theMaster volume a greenrectangle object named the Clipping plane was added in theabove defined interactive 3D scene to allow users to mimicthe clipping behavior in the Slavemodel see Figures 5(a) and5(b) Here a virtual control panel was designed separately tosupport the interactive behaviors of users using the Clippingplane (Figure 5(c))
During the initialization phase the Clipping plane coversthe 119883119884 plane in the world coordinate system (Figure 5(a))Its specific location could be accurately described by thefollowing two quantization parameters in the computergraphics method normal vector from the origin point tothe Clipping plane and the coordinate position of a pointon the Clipping plane Here its default normal vector was0 0 1 one point 119875 on the plane was on the position(0 0 0)
For intuitive operation of the user experience the interac-tive behaviors regarding the Clipping plane include rotationand panning and both can be operated by a virtual controlpanel However for the accurate description of the requestinstructions for remote volume rendering the two quantiza-tion parameters should be used
Therefore formula (1) was deduced to calculate thenormal vector from the angles of rotation
119899119900119903119898119886119897 V119890119888119905119900119903 = sin (120579119910) minus sin (120579119909)
sdot cos (120579119910) cos (120579119909) sdot cos (120579119910) (1)
Here 120579119909 120579119910 are the final rotation angles that rotate theClipping plane clockwise around the 119883-axis and 119884-axisrespectively
For example in Figure 5(b) a user rotated the Clippingplane 30 degrees clockwise around the 119884-axis and minus15degrees clockwise around the 119883-axis and then panned theClipping plane 95 pixels in the positive direction along the119885-axis While submitting for remote volume rendering 120579119909 =minus15 120579119910 = 30 the normal vector was 05 02241 08365according to formula (1) Meanwhile according to the char-acteristic of monolithic translation the one point 119875 on theClipping plane was moved to position (0 0 95) That wouldbe the two quantization parameters of request instructions forclipping
24 Remote Interaction between the Server and Client Theinteractive behaviors of the Slavemodel include viewing anglerequest and clipping request Moreover coloropacity request isoperated by the dropdown list of the preset schemes (detailedin Section 22) All of these requests would be submitted tothe server in the form of quantization parameters which arenamed as request instructions
However in addition to responding to the clientrsquos instruc-tions the server has to find an efficient way to manageand allocate system resources to enhance the robustness ofthe high frequency of interaction among multiple users forexample to reduce the probability of deadlock cumulativedelay and task-pipeline crossing
In our system we used relational database schema tomanage the request instructions from the client and to
6 International Journal of Telemedicine and Applications
Interactiveeect
Clippingplane
XY
Z
PX
Y
Z
(a) Initial state of Clipping plane
PX
Y
Z
(i) Rotated 30 degrees (ii) Further rotated minus15 degreesaround the X-axis
(ii) Further panning 95
P
XY
Z
P
XY
Z
Y-axisaround the Z-axisalong the
(b) Instances of interaction of Clipping plane (c) Control panelof Clipping plane
Figure 5 Design of the interaction of clipping to the Slavemodel and corresponding example
allocate the system resources for guiding 3D rendering Thecore tables and their data storage structure are designed inTable 1
According to Table 1 if a user chooses a group of SMIsfrom a list derived from the base table the server wouldquery the pipeline table whether the SMIs is serving theclient If No the SMIs would be loaded in memory anda task-pipeline between the user and SMIs would be builtThereafter the state of the pipeline table would be changed toldquoYesrdquoWhile being Yes the client would be permitted tomakethe next interactive request After that for each request fromthe client the servermust query the state of the behavior tablewhether the pipeline is ldquopendingrdquo or ldquoprocessingrdquo If No theserver would accept the request with a record of a ldquopendingrdquostate Additionally the listener on the server would real-timearrange these pending requests to the server rendering inqueue While a request is being rendered the correspondingstate of behavior table would be changed as ldquoprocessingrdquo toprevent malignant submission While the job is finished thestate would be changed to ldquofinishedrdquo and a 2D projectionimage would be generated and saved in the specified HTTPsharing path
The client adopted AJAX (Asynchronous JavaScript andXML) [31] to dynamically query the state of behavior tablefrom the server at a fixed frequency of 300ms While render-ing was finished the client would download and update the2D projection image via the HTML + js web page technique
3 System Description
31 System Architecture While Figure 1 depicts the Master-Slave dual-channel interaction mode it also presents theoverall architecture of the final platform Here we adoptedthe browserserver architecture for which the communi-cation with each other was via the HTTP protocol Theserver side consists of the Listener amp Request ManagementServer (in zone B) and Rendering Server (in zone C) Theformer accepts and manages the requests from the client
and guides the Rendering Server to an orderly responseThe latter renders the SMIs and generates a 2D projec-tion image for the client downloading via HTTP At thesame time the client side was designed by the standardof HTML5 [32] We used ltimggt an HTML image tagto display the 2D projection image in the form of pseudo3D and used AJAX + js technology to realize flicker-freepage updates and client RIA (rich Internet application)over diversified web browsers without installation More-over HTML5 includes the WebGL standard and the Slavemodel could run on any HTML5-supporting web browserswithout any web plug-ins or special software tools Thusthe whole platform system could run under mature networktechnologies
32 System Front-End System front-end is anHTML5-basedpure website that provides entrance to SMI access and 3Dpresentation The users require a username and password toenter the system or any experts could be invited to enter aspecific task-pipeline with one URL hyperlink from a helpseeker In brief using an HTML5-supporting web browserincluding the major versions of Firefox chrome Operasafari IE 10+ the 3D volume can be viewed and processedas in Figure 6 and the Supplementary Materials availableonline at httpsdoiorg10115520174074137 (1 Demon-stration of the operation procedure on laptop via websitemp4)
Here the interactive behaviors of the user include theviewing angle clipping and coloropacity transformationThefirst two items were operated on the Slave model which wasreconstructed on the client side While the user wants toobserve the final required image of high quality the userjust presses the submit button to submit the current behaviorrequest instruction to the server and waits for the final 2Dprojection image to be updated
Figure 6(a) presents the initial state of the Slave model(right) and 2D projection image (left) In Figure 6(b) mouseor hand gesture is used to operate the Slave model to
International Journal of Telemedicine and Applications 7
Table1Re
lationald
atas
torage
structureo
frequestbehaviorsfrom
clientand
resource
allocatio
non
thes
erver
Table
Field
Note
Basetable
DIC
OMs_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
grou
pof
SMIs
Torecord
theb
asicinform
ationof
SMIs
onthes
erverside
Storagep
ath
Thes
torage
path
ofSM
Ison
thes
erver
side
Pipelin
etable
pipelin
e_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
pipelin
efor
view
ingag
roup
ofSM
Isbetweenthec
lient
andserver
Basedon
basetableto
record
the
conn
ectedrelationshipbetweentheS
MIs
rend
eringin
mem
oryon
thes
ervera
ndremoteu
sero
nthec
lient
DIC
OMs_ID
FORE
IGNKE
Yof
basetableto
decla
rewhich
grou
pof
SMIsthec
lient
isview
ing
Client
Client
user
IPaddress
IPaddresso
fthe
committer
State
Whether
theS
MIsfin
ished
rend
eringin
mem
oryon
thes
erverYesallowthe
clienttonext
operation
Noprompt
for
waitin
g
Behavior
table
behavior_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofeach
behavior
requ
estevent
from
the
client
Basedon
pipelin
etable
torecord
operationrequ
estsfro
mthec
lient
user
andtheirp
rocesssta
te
pipelin
e_ID
FORE
IGNKE
Yof
pipelin
etable
todecla
rewhich
pipelin
eonview
ingand
processin
g
Behavior
type
Inclu
ding
view
inga
ngle
clippingor
coloropacity
transfo
rmation
Behavior-in
structio
n
From
thec
lienttotransfe
rthe
parametersinterm
softhe
correspo
nding
behavior
typerecordedin
theform
ofqu
antizationparameters
State
Pend
ingprocessin
gor
finish
ed
8 International Journal of Telemedicine and Applications
(a) Initial state (b) Viewing angle interaction
(c) Clipping interaction (d) Zooming and coloropacity interaction
Figure 6 Screenshot and operational demonstration
transform the viewing angle from Figure 6(a) to Figure 6(b)(red solid box) and then the submit button is pressed to sendthe behavior request instruction to the server to require anupdated projection image from the Master volume (yellowsolid box) In Figure 6(c) using virtual control panel (reddashed box) to operate Clipping plane after submitting thenew projection image was then updated (left) In Figure 6(d)the mouse wheel or hand gesture is used to zoom the Slavemodel and then the coloropacity theme is submitted via thedropdown list (bottom right corner) to guide remote volumerendering
4 Mode of Availability of the System
In this paper we designed aMaster-Slave dual-channel inter-action mode to improve a problem of remote visualizationinteraction of the volume reconstruction of SMIs whichusually have a huge data size and calculation cost Finallya visualization interaction platform has been built Particu-larly there is no special software and hardware requirementfor the clients and no special network environment isneeded
Tomeasure whether the effectiveness and performance toadapt to the mobile Internet application has characteristicsof lightweight three radiologists with more than 3 years ofexperience were invited to test the interactive experienceimaging quality and compatibility over diversified Internetdevices Moreover four junior students majoring in Biomed-ical Engineeringwere invited to test and quantize the networkload and response time
TestingData 3 groups of SMIs from the radiology departmentof Southwest Hospital of Chongqing include the following
(1) head and neck CT with 495 slices 512 lowast 512 2494MB(2) trunk CT with 609 slices 512 lowast 512 3065MB and (3)thorax andmandible CT with 500 slices 512lowast512 2516MBServer An IBM X3650M4 workstation (CPU 4 times E5-2603RAM 16GB OS Windows Server 2008) was connectedto the Internet via 10Mbps The 2D projection image wasset as the resolution of 512 lowast 512 with high quality JPEGcompression (level-8)
Client 4 Internet devices were located at the 3G or WIFImobile Internet with the server In Figure 7(a) there is aFounder E520 personal computer running on winXP withFirefox 46 web browser via 3G usb card In Figure 7(b) thereis an Acer E15 laptop running on win7 with chrome 43 webbrowser at WIFI In Figure 7(c) there is a HuaWei Honor6 smart phone running on EMUI 30 which is developedfrom Android 442 with the Opera 37 web browser at 3Gnet In Figure 7(d) there is an iPad mini running on IOS6with safari at WIFI All of them could be representative inthe mobile Internet Additionally each was limited to a 100-KBps wireless bandwidth by 360 security firewall for extremetest [33]
41 Interactive Experience Imaging Quality and Compatibil-ity To ensure the consistency of the context the testing data(1) were considered as the main object of reference Here thedemonstration and effect of the testing data (1) are shown inFigure 6
Figure 6 shows that this Slave model of the testing data(1) which was compressed to 104MB could be fit to thedemands of interactive navigation to accurately guide remotevolume rendering and shows that the projection image couldbe displayed to keep the original quality
International Journal of Telemedicine and Applications 9
(a) Personal computer (b) Laptop computer
(c) Smartphone (d) iPad
Figure 7 The system is accessed via four different devices and web browsers under extreme bandwidth
For further validation three radiologists were invited totest the platform with the similar behaviors in Figure 6 Thisjob was repeated to operate three groups of different testingdata through different web browsers and Internet devicesthat were all aforementioned Additionally a five-pointscale was designed to respond to the radiologistsrsquo requests(Table 2)
Table 2 presents a very positive evaluation from the statis-tical result of the imaging quality and interactive experienceand all three radiologists expressed that is an interestingway whether as a gesture-based or mouse-based interactionNo installations were required from the client and goodcompatibility was obtained (same result is shown in Table 3)However one radiologist thought the default Slave modelof the testing data (1) with a skeleton model of 104MB(Figure 3(d2)) was a little rough He suggested to use theskeleton model of 526MB or skin model of 913 KB (Figures3(c1) and 3(c2)) instead of the 104 model both of which arethe alternatives from the mutual interface for the client asdetailed in Section 231 Additionally all three radiologistsclaimed that the response time for final high quality image isalways about 1 second that is consistent with the test resultin Section 42 However they did accept the performancebecause they could directly operate the Slavemodel smoothlywithout any delay on client side thus a short wait for finalhigh quality image had little influence Moreover one of theradiologists specifically expressed the native application of3D rendering on desktop PC also needs 1 or 2 secondsrsquo delayfor final image this performance of the solution running onsmart phone is similar to the native application on PC and isvery good
42 Response Time and Network Load between the Clientand Server The response time can be defined as the timeinterval between the user sending a behavior request tothe system and web browser loading the response resultscompletely The network load can be defined as the amountof data transmission of each interactive behavior Four juniorstudents majoring in Biomedical Engineering were invitedto simultaneously test the platform for one hour During thetest all of the students respectively viewed and processed thetesting data (1) through the platform and performed similarbehaviors to those detailed in Figure 6This test was repeatedby each student with different web browsers and Internetdevices all of which were aforementioned Here we trainedthe students to record the real-time network traffic by the 360security firewall and the time interval by the script log Theresults are showed in Table 3
In Table 3 for each device the amount of data trans-mission and response time of the first loading are muchhigher than the behavior request and idle status because thefirst loading asks the client to download the necessary Slavemodel The response time of the first loading is lt12 s at 100-KBps client bandwidth which is similar to the total timespent on Slavemodel downloading In fact the response timeof first loading also includes the cost time of SMI loadingand rendering on the server side but the server monitorpresents the cost time as lt7 s because SMI loading and Slavemodel downloading are simultaneous and this cost could beneglected
After the first loading the network load of the behaviorrequest for final image is spent on request instructions sub-mission final projection image downloading and the serverprocessing status querying at a fixed frequency of 300ms
10 International Journal of Telemedicine and Applications
Table 2 Statistic results of the five-point scale concerning the interactive experience and imaging quality
Focus Subject Votes from five-point scaleVeryagree Agree Uncertainty Disagree Very
disagree
Interactive experienceMouse-basedGesture-based 998771
The operation of viewing and processing isaccurate
3998771 3
Simple and easy to use without training 2 1998771 3
A good user experience similar to nativeapplication
2 1998771 2 1
Imaging quality
Slavemodel could meet demands of navigation 2 12D image fromMaster volume could provide highquality reconstruction result via utilizing thedisplay capability of Internet terminals
3
Others Latency time for final high quality images 3
Table 3 Average response time and amount of data transmission from the different behaviors and devices at 100-KBps
Devices
First loading (averagevalue)
Behavior request for finalimage (average value)
Idle status ofbandwidthoccupation(KBps)
Frames PerSecond (FPS)Amount of
data trans (KB)Response time(seconds)
Amount ofdata trans (KB)
Response time(seconds)
PC (3G net) 108544 1123 4710 112 023 48ndash60Laptop (WIFI) 106496 1051 4608 089 031 48ndash60iPad mini (WIFI) 107941 1093 4403 095 035 50ndash60Smartphone (3G net) 107502 1072 3994 102 028 30ndash40
where the projection image downloading is the main partThe results present a complete amount of data transmission ofeach behavior request is close to 50KB but the correspondingresponse time for final image is approximately 1 s in 100KBpsThat is because besides the network load the rest of timeis spent on server processing and time interval of querying(300ms) The short delay could be acceptable for practicalapplications as detailed in Section 41
While in the idle status including the userrsquos interactivebehavior (operation) on the Slave model the bandwidthoccupation is almost negligible Moreover there are no sig-nificant differences among diversified devices and networksat 100-KBps client bandwidth
Additionally the FPS of Slave model rendering on theclient side is measured via statsminjs which is a librarypackaged from WebGL When the client rendering is com-pleted FPS could remain at the upper limit
5 Discussion and Comparison
Due to the nonprivate network environment and diversifiedpurposes application software on one device must competefor limited resources of the network and computing Thuslightweight which requires lower resources of the networkand computing and avoids unnecessary installation could bean important characteristic of mobile Internet applicationAdditionally to avoid compatibility problems caused by dif-ferent operating systems and client hardware characteristics
using HTML5 and related web technologies running onmajor web browsers may be a choice to adapt to cross-platform deployment and the characteristics of lightweight[21]
Therefore this article described research on a methodfor the Master-Slave dual-channel interaction mode whichis based on the HTML5 standard (also a wider web com-patibility standard of WebGL) to permit users viewing andprocessing million-megabyte-class SMIs via different clientdevices (without hardware or operating system constraints)over a pure web page (without installing special software) andin a diversified network environment
Compared with the traditional PACS mode of ldquoimagecompression transmission local reconstructionrdquo becauseour method does not transfer original SMIs to the clientit neither needs to wait for a long time to download norrequests special demands of the software environment onthe client device Moreover compared with the state ofthe art of local volume rendering based on WebGL andHTML5 technology [7ndash9] our methods have no specialdemands of client hardware for computation that may reducebattery consumption [9] and enhance the compatibility ofthe client devices our methods may be more fit for mobileInternet application Additionally the client does not accessthe original SMIs which could be much safer
Compared with popular solutions based on absolutesurface rendering technology [11 12 14] our method haslocally used this technique to only support light Slavemodel
International Journal of Telemedicine and Applications 11
for interactive navigation therefore the demands of thenetwork bandwidth and device rendering capability arealmost negligible in ourmethodMoreover the final requiredimage of high quality is a projection drawing from volumerendering technology indicating that our method could beconvenient for users to view the internal anatomy structureand complex organization relationship however surfacerendering technology could not support these characteristicsof volume
Compared with solutions based on absolute remote vol-ume rendering methods our method adopted the Master-Slave dual-channel interaction mode to strengthen the userexperience which is an entirely different approach from themode of ldquocontinuous request continuous responserdquo of studiesof [9] Although studies of [17ndash19] proposed several methodsto improve the mode to reduce the network load includingvideo-compressed transmission [17] tile-based transmission[18] and variable resolution transmission [19] the pressureof server load and real-time data stream increased We usedthe same SMI data (testing data (1)) the same quality offinal required image and the same server configurationand volume rendering algorithm in Section 4 to test themethods in [17ndash19] For quantitative test we formulated aninteractive trajectory with fixed 30 request points theoreti-cally assuming 30 rendering requests triggered per 1 secondthere should be 30 fps render effects On this basis [17]adopted 512 lowast 512 mpeg-4 compression for transmission[18] adopted 8 lowast 8 blocks to segment each 512 lowast 512 frameand transfer the blocks which are different from previousframe in the process of interaction and [19] adopted 64 lowast64 resolution transmission in the process of interaction but512 lowast 512 for end frame The test was repeated for 5 timesunder the condition of one single client monopolizing thebandwidth and server under ideal WIFI environment Theresults showed that the total amount of data transmissionfor each interaction test of [17ndash19] was 2980 KB 6036 KBand 3713 KB respectively the average response time of firstframe was 110 seconds 059 seconds and 052 seconds andthe average cumulative response time of end frame (30thframe which is the final required image) was 2131 seconds1761 seconds and 1601 seconds The response time containsthe factors of server continuous rendering response servercontinuous frame-encoding and so forth (more than 70server CPU utilization during interaction) but the networkfactors could be tentatively neglected under the ideal single-user WIFI environment Assuming that we can improvethe serverrsquos processing ability to completely eliminate theserver response delay (perfect 30 fps interaction effect) thetheoretical value of the bandwidth could be 2980 KBps6036 KBps and 3713 KBps respectively In summary theFPS value one of the most important parameters of the 3Dinteractive experience of the studies of [17ndash19]was correlatedreciprocally with the number of simultaneous tasks on theserver side (server task-pipelines) server processing abilityand network bandwidth and an obvious delay of interactionwas clearly presented By contrast all of the Internet deviceschosen to test our method could provide an FPS value at30ndash60 and the FPS valuewas not correlatedwith server tasksserver processing ability and network bandwidth but with
the device GPU (Graphics Processing Unit) Undoubtedlyalmost all GPUs of Internet devices in the market could meetthe demands of our light Slave model rendering Moreoverthe peak value of the bandwidth under the behavior requestfor final image was less than 50KBps and less than 1 KBpsat idle (including the operation of Slave model) thus bothwere far below those tests of methods of [17ndash19] Thusour method could be more suitable for a complex mobileInternet environment Moreover the complex compressiontransmission methods of [17ndash19] were built on customizedsoftware of plug-ins for the client which is hard to applyto pure web pages and falls short of wide compatibility Bycontrast that is our methodrsquos advantage
For the server load because the Master volume in ourmethod only responds to the final behavior request (instruc-tions) instead of the continuous request of the studies of[9 17ndash19] the test showed that server CPUutilizationwas lessthan 20 at idle (including the operation of Slavemodel) andless than 30 for responding to behavior request thus it wasfar below of those tests ofmethods of [17ndash19] with nearly a fullload during all the operation time under the same conditionsof SMI data quality of final required image server config-uration and volume rendering algorithm Theoretically ourplatform can respond to more independent tasks (pipelines)at the same time Table 2 shows that while the server wasrunning four independent tasks simultaneously the responsetime of our method remained at approximately 1 s in the 100-KBps client bandwidth much better than [17ndash19] which waswith more than 16 secondsrsquo delay of end frame in the singletask status In fact because user viewing and processing ofthe 3D medical volume occurred via interaction on the Slavemodel the delay experience wasmuch fewer than that in [17ndash19] as detailed in Section 41That indicates ourmethod couldhave some more advantages including much less networkload and server load to serve much more independent tasksand users
In addition [17ndash19] did not propose an effective inter-active method to clip the 3D volume what our method hasaccomplished
6 Conclusion
This paper proposed aMaster-Slave dual-channel interactionmode which absorbed the advantages of the surface ren-dering of WebGL and remote volume rendering to builda visualization interaction platform to rapidly view andprocess 3Dmedical volume reconstruction in a pure webpageenvironment This platform could be used in a diversifiednetwork environment (100-KBps bandwidth for extreme)and diversified client devices (no special hardware and oper-ating system constraints) without any special installation butwith a good user experience similar to native applicationThese features could serve authorized users to convenientlyaccess SMIs and present 3D visualization anywhere andcould be a technological basis of online communicationamong doctors and patients [34] Or at least it could helpclinicians to arbitrarily acquire 3D image according to theirclinical needs instead of the static pictures that radiologistsdrafted For example clinicians could conveniently observe
12 International Journal of Telemedicine and Applications
the complicated relationship of skull base bone and bloodvessel via this platform (Figure 6(d)) in the past this pieceof information was provided by a few static images generatedfrom radiologistrsquos understanding
Additionally this solution is designed to be followed asanHTML5 standard interface indicating any authorized usercould rapidly access the application via one URL hyperlinkand the platform could be easily integrated into othertelemedical systems seamlessly as a third-party applicationfor example the Supplementary Materials (2 Demonstrationof the operation procedure on laptop linked by emailmp4and 3 Demonstration of the operation procedure on iPadvia websitemp4) These features may play an assistant roleto improve work patterns and extend the range of applicationof traditional medical image systems For example it can beapplied to regional health telemedicine andmedical imagingeducation at a low cost because there is no special cost ofdeploy and upgrade of software network and hardware onthe client side And more importantly a teacher can vividlyshow the students the image characteristics of the diseaseanywhere a clinician can intuitively introduce disease to thepatients anywhere and ask expertsrsquo help anywhere That maypromote the development of mobile medical technology to acertain degree
Conflicts of Interest
There are no conflicts of interest declared
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China (no 81171866) the National Key BasicResearch Program of China (no 2014CB541602) and theSocial Livelihood Science andTechnology Innovation SpecialProject of CSTC (no cstc2015shmszx120002)
References
[1] Q Zhang M Alexander and L Ryner ldquoSynchronized 2D3Doptical mapping for interactive exploration and real-time visu-alization of multi-function neurological imagesrdquo ComputerizedMedical Imaging and Graphics vol 37 no 7-8 pp 552ndash567 2013
[2] S Zimeras and L G Gortzis ldquoInteractive tele-radiological seg-mentation systems for treatment and diagnosisrdquo InternationalJournal of Telemedicine and Applications vol 2012 Article ID713739 p 15 2012
[3] S E Mahmoudi A Akhondi-Asl R Rahmani et al ldquoWeb-based interactive 2D3D medical image processing andvisualization softwarerdquo Computer Methods and Programs inBiomedicine vol 98 no 2 pp 172ndash182 2010
[4] L Faggioni E Neri I Bargellini et al ldquoIPad-based primary2D reading of CT angiography examinations of patients withsuspected acute gastrointestinal bleeding preliminary experi-ence 1Lrdquo British Journal of Radiology vol 88 no 1047 ArticleID 20140477 2015
[5] J B Park H J Choi J H Lee and B S Kang ldquoAn assessment ofthe iPad 2 as a CT teleradiology tool using brain CT with sub-tle intracranial hemorrhage under conventional illuminationrdquoJournal of Digital Imaging vol 26 no 4 pp 683ndash690 2013
[6] P T Johnson S L Zimmerman D Heath et al ldquoThe iPad asa mobile device for CT display and interpretation diagnosticaccuracy for identification of pulmonary embolismrdquo EmergencyRadiology vol 19 no 4 pp 323ndash327 2012
[7] J M Noguera and J R Jimenez ldquoVisualization of very large 3Dvolumes on mobile devices and WebGLrdquo in Proceedings of the20th International Conference in Central Europe on ComputerGraphics Visualization and Computer Vision WSCG 2012 pp105ndash112 June 2012
[8] J Congote A Segura L Kabongo et al ldquoInteractive visu-alization of volumetric data with WebGL in real-timerdquo inProceedings of the 16th International Conference on 3D WebTechnology (Web3D rsquo11) pp 137ndash145 June 2011
[9] A Schiewe M Anstoots and J Kruger ldquoState of the art inmobile volume rendering on iOS devicesrdquo in Proceedings of theEurographics Conference on Visualization (EuroVis rsquo15) 2015
[10] A Evans M Romeo A Bahrehmand et al ldquo3D graphics on theweb a surveyrdquo Computers amp Graphics vol 41 no 1 pp 43ndash612014
[11] M Callieri R M Andrei M D Benedetto et al ldquoVisualizationmethods formolecular studies on theweb platformrdquo inProceed-ings of the 15th international conference on web 3D technology(Web3D rsquo10) pp 117ndash126 July 2010
[12] J Jimenez A M Lopez J Cruz et al ldquoA Web platform forthe interactive visualization and analysis of the 3D fractaldimension of MRI datardquo Journal of Biomedical Informatics vol51 pp 176ndash190 2014
[13] WebGL httpswwwkhronosorgwebgl[14] C R Butson G Tamm S Jain T Fogal and J Kruger
ldquoEvaluation of interactive visualization on mobile computingplatforms for selection of deep brain stimulation parametersrdquoIEEE Transactions on Visualization and Computer Graphics vol19 no 1 pp 108ndash117 2013
[15] M Levoy ldquoDisplay of surfaces from volume datardquo IEEE Com-puter Graphics and Applications vol 8 no 3 pp 29ndash37 1988
[16] E Kotter T Baumann D Jager and M Langer ldquoTechnologiesfor image distribution in hospitalsrdquo European Radiology vol 16no 6 pp 1270ndash1279 2006
[17] S Park W Kim and I Ihm ldquoMobile collaborative medical dis-play systemrdquo Computer Methods and Programs in Biomedicinevol 89 no 3 pp 248ndash260 2008
[18] J R Mitchell P Sharma J Modi et al ldquoA smartphone client-server teleradiology system for primary diagnosis of acutestrokerdquo Journal of Medical Internet Research vol 13 no 2 p e312011
[19] T Hachaj ldquoReal time exploration and management of largemedical volumetric datasets on small mobile devicesmdasheval-uation of remote volume rendering approachrdquo InternationalJournal of Information Management vol 34 no 3 pp 336ndash3432014
[20] P Quax J Liesenborgs A Barzan et al ldquoRemote renderingsolutions using web technologiesrdquoMultimedia Tools and Appli-cations vol 75 no 8 pp 4383ndash4410 2016
[21] D Preuveneers Y Berbers and W Joosen ldquoThe future ofmobile E-health application development exploring HTML5for context-aware diabetes monitoringrdquo Procedia ComputerScience vol 21 pp 351ndash359 2013
[22] J M Noguera J J Jimenez and M C Osuna-Perez ldquoDevel-opment and evaluation of a 3D mobile application for learningmanual therapy in the physiotherapy laboratoryrdquo Computersand Education vol 69 pp 96ndash108 2013
International Journal of Telemedicine and Applications 13
[23] Y S Kang ldquoVolume rendering overviewrdquo inGPU Programmingand Cg Language Primer chapter 14 pp 154-155 1st edition2009
[24] Visualization Toolkit (VTK) httpwwwvtkorg[25] vtkPiecewiseFunction Class Reference of VTK httpwwwvtk
orgdocnightlyhtmlclassvtkPiecewiseFunctionhtml[26] vtkColorTransferFunction Class Reference of VTK http
wwwvtkorgdocnightlyhtmlannotatedhtml[27] vtkPlane Class Reference of VTK httpwwwvtkorgdoc
nightlyhtmlclassvtkPlanehtml[28] L Ma D-X Zhao and Z-Z Yang ldquoA software tool for
visualization of molecular face (VMF) by improving marchingcubes algorithmrdquo Computational and Theoretical Chemistryvol 1028 pp 34ndash45 2014
[29] G F Wang J Liu and W Liu ldquoDIB display and window trans-formation technology for DICOM medical imagesrdquo ChineseJournal of Medical Device vol 18 no 8 pp 1ndash5 2005
[30] L Qiao X Chen L X Yang et al ldquoDevelopment of medi-cal image three-dimensional-visualization platform based onhome network and pure web conditionsrdquo Beijing BiomedicalEngineering vol 34 no 3 pp 229-233286 2015
[31] M Ying and J Miller ldquoRefactoring legacy AJAX applications toimprove the efficiency of the data exchange componentrdquo Journalof Systems and Software vol 86 no 1 pp 72ndash88 2013
[32] HTML5 httpwwww3orgTR2012CR-html5-20121217[33] QIHU 360 SOFTWARE CO LIMITED 360 Internet Security
httpwww360safecomInternet-securityhtml[34] L Qiao Y Li X Chen et al ldquoMedical high-resolution image
sharing and electronic whiteboard system a pure-web-basedsystem for accessing and discussing lossless original images intelemedicinerdquoComputerMethods and Programs in Biomedicinevol 121 no 2 pp 77ndash91 2015
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
2 International Journal of Telemedicine and Applications
interested area the client user cannot arbitrarily view theinternal anatomy structure or analyze the complex orga-nization relationship Therefore surface rendering technol-ogy has been mainly used for special disease with specificrequirements [11 12 14] and it could not adapt to thecustom behavior of the client user in mobile Internet verywell
Volume rendering describes a wide range of techniques togenerate images from 3D scalar data [15] It is very convenientto view the internal anatomy structure and describe thecomplex organizational relationship through user adjustmentof the parameters of color and opacity on each voxel [16]However volume rendering requires a complete computationof the original volumetric data for each userrsquos behavior eventthe cost of computation being much higher than that for sur-face rendering Thus volume rendering technology is oftendeployed on a dedicated image workstation and in a privatenetwork environment where the scope of application is lim-ited Some low-level graphics APIs such as OpenGL ES 20have been integrated into WebGL for Internet applicationwhich could be theoretically enabled to support GPU-baseddirect volume ray-casting implementations or support someslice-based sampling techniques to get close to the imagingresult of ray-casting-based implementations Neverthelessthe imaging quality and efficiency is very limited not justbecause of performance of common GPU but also becauseof the limited performance of web pages [7ndash9] Further-more there is still high requirement to network for originalvolumetric data transmission Currently there is a remotevolume rendering scheme to solve the limit This schemedeploys the volume rendering job on the server side to a real-time response to behavior instructions from the client andtransmits the 2D projection image from volume renderingto the client display where the client is only an interface forsending instructions and displaying results [9 17ndash20] Forexample the user slides the mouse from point A to point B toexpress an operation of 3D rotation on the client side and theserver just calculates the final viewing angle from A to B andsends the final result to the user Because there is no originaldata downloading and no rendering cost on the client sidethe network load and computing pressure of the client couldbe significantly reduced However this interaction model ofldquothe client sends one operation request the server replies oneresult at a timerdquomay cause a bad user experience Specificallyit may lose the value of Frames Per Second (FPS) one of themost important parameters of the 3D interactive experienceTherefore all of [9 17ndash20] discretized the trajectory of onesingle operation request from the client into several requestpoints to request continuous responses from the server Forexample if the path between mouse points A and B wereconverted into several separated remote volume renderingrequests from the client the sever would respond to eachrequest and send continuous 2D projection images to theclient Thus the value of FPS could be increased howeverit may lead to a huge pressure on the server load (responseto a large amount of requests) and network load (continuousdata stream) Additionally [9 17ndash19] ask the client to installcustomized software or plug-ins which may lead to a lack ofcompatibility Therefore existing remote volume rendering
methods do not adapt to themobile Internet application char-acteristics of lightweight [21] very well which ask for lowerresources of network and computing and avoid unnecessaryinstallation
In this study we plan to combine the advantage of surfacerendering technology based on WebGL with the advantageof remote volume rendering to design a solution for viewingand processing 3D medical volume reconstruction in a purewebpage environment Finally a visualization interactionplatform would be built This platform could be used any-where in a diversified network environment and diversifiedclient devices with a good user experience similar to thenative application
2 Computational Methods and Theory
The core concept includes the Slave model and Mastervolume The Slave model is a contour model rendered bysurface rendering technology running on WebGL as a 3Dinteractive interface on the client side The Master volumeis a volumetric data rendering from original SMIs on theserver side The final required image of high quality is aprojection drawing from theMaster volume according to thebehavior instructions from the Slave model This solutioncould be named the Master-Slave dual-channel interactionmode and is expected to enhance the user experience similarto the native application in a pure web page to reduce thepressure of the server load and network load and to meetthe requirement of lightweight volumetric data sharing in amobile Internet application
21 Design of the Master-Slave Dual-Channel InteractionMode The architecture of this solution constitutes the fol-lowing two parts the server side and client side The serverside renders theMaster volume to generate the final requiredimage of high quality while the client side provides the Slavemodelrsquos interactive navigation
The detailed procedures are as follows (schematic dia-gram is shown in Figure 1)
On the server side in zone A groups of SMIs couldbe uploaded to the server through the network interface ofPACS or removable storage medium Thereafter the relevantSlave models could be automatically pregenerated by roughcontour extraction The listener in zone B accepts theuserrsquos web behavior request (instructions) in real-time andguides zone C to load and render the relevant SMIs inmemory which is defined as the Master volume generatinga 2D projection image from the rendered result for clientdownloading
On the client side while the user wants to view andprocess a group of SMIs the client could download thecorresponding Slavemodel from the server and reconstruct itlocally in zoneD meanwhile request the server to load andrender relevant SMIs in memory Thereafter in zone D theuser could view and process the local Slavemodel arbitrarilyand submit the final behavior request (instructions) to theserver to acquire the required projection image from theMaster volume in zoneB C E
International Journal of Telemedicine and Applications 3
Interactivebehaviors
Sectional medical images (SMIs) amp Slave model
Service
N
M
Any device amp OS amp browser
e screen eect
2D imageMaster volume
Listener
Projection
Slave model
Internet
Client
Volume rendering
Behaviorrequest
Request Mgmt
Loading in memory
1
2
3
85
86
middot middot middot
middot middot middot
middot middot middot
12
3
85
86
12
3
8586
middot middot middot
①
④
⑤
③
②
Figure 1 Architecture of the pure web-based solution via theMaster-Slave dual-channel interaction mode
Here the client could be a pure web browser running ondiversified devices supporting the WebGL technique Thatmeans there is no installation for the client
22 Rendering of the Master Volume We adopted the ray-casting algorithm [22] to reconstruct the Master volume ofSMIsThe algorithm is a direct volume renderingmethod thathas been widely used for radiodiagnosis but requires a highcomputing performance of the device First we transformedthe SMIs into volume voxels in 3D space (Figure 2(a)) andthen simulated rays to cast each voxel with a customizedcolor and opacity [23] into 2D space (Figure 2(b)) Thismethod could be used to describe a complex organizationrelationship Furthermore it could display internal anatomystructure via removing (clipping) part of the voxels thatsimulate a userrsquos behavior of dissection (Figure 2(c))
Here we directly used the VTK (Visualization Toolkit)library [24] to realize the ray-casting algorithm For cus-tomized color and opacity several frequently used schemescould be deployed on the server side in advance a dropdownlist was designed on the user interface to correspond to eachscheme for remote volume rendering [25 26] For clippingVTK provides a class of vtkPlane to build a virtual Clippingplane to guide the voxel removal [27] It means the clippingjob only needs the specific location of the virtual Clippingplane in the 3D coordinate systemThe specific location couldbe described by the following two parameters in the computergraphics method normal vector from the origin point to theplane and the coordinate position of a point on the Clippingplane both from the client submission
23 Generation and Interaction of the Slave Model The Slavemodel was pregenerated on the server side and was designedto be a navigational tool to interact with theWeb page Whileclients want to view and process theMaster volume the Slavemodel could be downloaded to the client and reconstructedon the client side
231 Generation and Slimming of the SlaveModel To interactwith mobile Internet the data size of the model should beas small as possible (much slimmer) while the clarity ofthe contour is acceptable However using the traditionalcompression-decompression algorithm may ask the client toinstall additional software Hence the steps in the generationof a smaller Slavemodel can be designed as follows (Figure 3)Here we used a group of SMIs of the head and neck CT asan example and the resolution of voxels is 512 times 512 times 495249MB (testing data (1) in Section 4)
Step 1 (resampling the original SMIs for first slimming) Loadthe corresponding SMIs into memory shrink each sectionalimage to a standard size of 32 lowast 32 via the resampling of the2D space (Figures 3(a) and 3(b))
Step 2 (building the 3D geometrical contour model) Amarching cubes algorithm [28] one of the most commonmethods of surface rendering was used to extract theinterested surface contour of the anatomical representationfrom resampling SMIs into a 3D geometrical contour modelHere according to the feature of the CT grayscale [29] weset a particular contour value at minus800 or 200 that couldseparate the skin or skeleton of the body from the resampling
4 International Journal of Telemedicine and Applications
(a) Voxels in volume data (b) Pseudo color amp opacity for each voxel (c) Clipping
Figure 2 Schematic diagram of the direct volume rendering method
Size YSize X
Size ZSize Z
Size XSize Y
Y Y
X X
Z Z
(a) Original SMIs
Step 1 Resampling the original SMIs for rst compression
Shrink each slice
(b) Resampling amp strech
(c1) Contour value minus800
(c2) Contour value 200Size 526MBStep 2 Building
3D contour model
(d1) Polygons 50 Polygons 10
(d2) Polygons 50 Polygons 10
Step 3 Reduce the triangles
Step 4Saved asvtk le
(eg 512 times 512 times 495 249 MB) 311MBSize 104MBSize
333MBSize 913MBSize674MBSize
2 times 32 times 495)(eg 3
Figure 3 Schematic diagram of the generation and compression of the Slavemodel
SMIs respectively to acquire a skin model or skeleton model(Figures 3(c1) and 3(c2)) In this case the data size of the skinmodel and skeleton model could be reduced to 674MB and526MB respectively
Step 3 (reducing the total number of polygons for secondslimming) The 3D geometrical contour model from Step 2is constitutive of polygons and we could reduce the totalnumber of polygons to further reduce the data size In thiscase the number was reduced to 50 and 10 and thedata sizes were reduced to approximately 3MB and 1MBrespectively (Figures 3(d1) and 3(d2))
Step 4 (saved as vtk model file) The slimmed model wouldbe saved in a file of the vtk format which is a storage formatof the geometrical model supported by WebGL and thenstored in the same directory with original SMIs
The result of Figures 3(d1) and 3(d2) preliminarily showedthat the Slave model whose data size is approximately 1MBstill could be fit for interactive navigation (more experimentsdetailed in Section 41) Moreover an additional mutualinterface of the client was designed as a remote parametersetting of the contour value and the total number of polygonsto permit users to customize the Slavemodel as desired
232 Design and Quantization of the Interactive Behaviors inthe Slave Model For a friendly interactive experience on theclient side a simple style of interactive behaviors of the clientand corresponding quantitative methods should be designed
to guide remote volume rendering including the viewingangle request and clipping request
For the Viewing Angle Request The interaction of viewingangle includes rotation flipping zooming and panning [30]We defined the interactive 3D scene including only one Slavemodel and one camera and customized an interactive modeof the camera in VR technology to meet our requirementFigure 4(a) shows the principle of the interaction of theviewing angle between the camera and Slave model Welimited the lens of the camera to always be taken on point1198751 (0 0 0) which was the origin of the world coordinatesystem The interactive behaviors of the viewing angle werecarried out by only changing the camera position exceptpanning For panning it was carried out by changing theSlavemodel position see examples in Figure 4(b)
According to the definition while sending requestinstructions to the server for remote volume rendering theinstructions only include three quantization parameters fromthe client to accurately describe the interactive behavior ofthe viewing angle request to the Slavemodel The three quan-tization parameters include the following attribute view updirection attribute position of camera and attribute positionof the center point of the Slavemodel
Here attribute view up direction of camera indicates thedirection of the top of the camera For example (0 minus1 0)a value of the View Up direction indicates the camera isperpendicular to the XZ plane in the world coordinatesystem and the top of the camera is in the opposite direction
International Journal of Telemedicine and Applications 5
(a) Principle of interaction between camera and Slavemodel
(i) Original camera view
(iii) Zoom via shorteningthe distance from CaA to P1
(iv) Flip from CaA to CaB (v) Panning via movingthe center point of Slavemodel from P1 to P2
(ii) Rotate around axis CaA-P1 (Y-axis)
(b) Example of interactions of viewing angle request
Figure 4 Principle of the interaction of the viewing angle between the camera and Slavemodel and the corresponding example
along the 119885-axis (see the relationship between camera CaAand Slave model in Figure 4(a) and corresponding example(i) original camera view in Figure 4(b))
In fact WebGL has provided some interaction methodsof the moving camera to view 3D objects via mouse- andgesture-based interactive behaviors which were packagedinto a library of TrackballControlsjs for a web page scriptWe could directly utilize the js library to acquire the threequantization parameters However we need to rewrite thejs library to restrict and customize the interactive modeof the camera that we have defined to ensure the consis-tency of the interactive rule between the Master and Slaveside
For the Clipping Request For arbitrarily clipping and viewingthe internal anatomy structure of theMaster volume a greenrectangle object named the Clipping plane was added in theabove defined interactive 3D scene to allow users to mimicthe clipping behavior in the Slavemodel see Figures 5(a) and5(b) Here a virtual control panel was designed separately tosupport the interactive behaviors of users using the Clippingplane (Figure 5(c))
During the initialization phase the Clipping plane coversthe 119883119884 plane in the world coordinate system (Figure 5(a))Its specific location could be accurately described by thefollowing two quantization parameters in the computergraphics method normal vector from the origin point tothe Clipping plane and the coordinate position of a pointon the Clipping plane Here its default normal vector was0 0 1 one point 119875 on the plane was on the position(0 0 0)
For intuitive operation of the user experience the interac-tive behaviors regarding the Clipping plane include rotationand panning and both can be operated by a virtual controlpanel However for the accurate description of the requestinstructions for remote volume rendering the two quantiza-tion parameters should be used
Therefore formula (1) was deduced to calculate thenormal vector from the angles of rotation
119899119900119903119898119886119897 V119890119888119905119900119903 = sin (120579119910) minus sin (120579119909)
sdot cos (120579119910) cos (120579119909) sdot cos (120579119910) (1)
Here 120579119909 120579119910 are the final rotation angles that rotate theClipping plane clockwise around the 119883-axis and 119884-axisrespectively
For example in Figure 5(b) a user rotated the Clippingplane 30 degrees clockwise around the 119884-axis and minus15degrees clockwise around the 119883-axis and then panned theClipping plane 95 pixels in the positive direction along the119885-axis While submitting for remote volume rendering 120579119909 =minus15 120579119910 = 30 the normal vector was 05 02241 08365according to formula (1) Meanwhile according to the char-acteristic of monolithic translation the one point 119875 on theClipping plane was moved to position (0 0 95) That wouldbe the two quantization parameters of request instructions forclipping
24 Remote Interaction between the Server and Client Theinteractive behaviors of the Slavemodel include viewing anglerequest and clipping request Moreover coloropacity request isoperated by the dropdown list of the preset schemes (detailedin Section 22) All of these requests would be submitted tothe server in the form of quantization parameters which arenamed as request instructions
However in addition to responding to the clientrsquos instruc-tions the server has to find an efficient way to manageand allocate system resources to enhance the robustness ofthe high frequency of interaction among multiple users forexample to reduce the probability of deadlock cumulativedelay and task-pipeline crossing
In our system we used relational database schema tomanage the request instructions from the client and to
6 International Journal of Telemedicine and Applications
Interactiveeect
Clippingplane
XY
Z
PX
Y
Z
(a) Initial state of Clipping plane
PX
Y
Z
(i) Rotated 30 degrees (ii) Further rotated minus15 degreesaround the X-axis
(ii) Further panning 95
P
XY
Z
P
XY
Z
Y-axisaround the Z-axisalong the
(b) Instances of interaction of Clipping plane (c) Control panelof Clipping plane
Figure 5 Design of the interaction of clipping to the Slavemodel and corresponding example
allocate the system resources for guiding 3D rendering Thecore tables and their data storage structure are designed inTable 1
According to Table 1 if a user chooses a group of SMIsfrom a list derived from the base table the server wouldquery the pipeline table whether the SMIs is serving theclient If No the SMIs would be loaded in memory anda task-pipeline between the user and SMIs would be builtThereafter the state of the pipeline table would be changed toldquoYesrdquoWhile being Yes the client would be permitted tomakethe next interactive request After that for each request fromthe client the servermust query the state of the behavior tablewhether the pipeline is ldquopendingrdquo or ldquoprocessingrdquo If No theserver would accept the request with a record of a ldquopendingrdquostate Additionally the listener on the server would real-timearrange these pending requests to the server rendering inqueue While a request is being rendered the correspondingstate of behavior table would be changed as ldquoprocessingrdquo toprevent malignant submission While the job is finished thestate would be changed to ldquofinishedrdquo and a 2D projectionimage would be generated and saved in the specified HTTPsharing path
The client adopted AJAX (Asynchronous JavaScript andXML) [31] to dynamically query the state of behavior tablefrom the server at a fixed frequency of 300ms While render-ing was finished the client would download and update the2D projection image via the HTML + js web page technique
3 System Description
31 System Architecture While Figure 1 depicts the Master-Slave dual-channel interaction mode it also presents theoverall architecture of the final platform Here we adoptedthe browserserver architecture for which the communi-cation with each other was via the HTTP protocol Theserver side consists of the Listener amp Request ManagementServer (in zone B) and Rendering Server (in zone C) Theformer accepts and manages the requests from the client
and guides the Rendering Server to an orderly responseThe latter renders the SMIs and generates a 2D projec-tion image for the client downloading via HTTP At thesame time the client side was designed by the standardof HTML5 [32] We used ltimggt an HTML image tagto display the 2D projection image in the form of pseudo3D and used AJAX + js technology to realize flicker-freepage updates and client RIA (rich Internet application)over diversified web browsers without installation More-over HTML5 includes the WebGL standard and the Slavemodel could run on any HTML5-supporting web browserswithout any web plug-ins or special software tools Thusthe whole platform system could run under mature networktechnologies
32 System Front-End System front-end is anHTML5-basedpure website that provides entrance to SMI access and 3Dpresentation The users require a username and password toenter the system or any experts could be invited to enter aspecific task-pipeline with one URL hyperlink from a helpseeker In brief using an HTML5-supporting web browserincluding the major versions of Firefox chrome Operasafari IE 10+ the 3D volume can be viewed and processedas in Figure 6 and the Supplementary Materials availableonline at httpsdoiorg10115520174074137 (1 Demon-stration of the operation procedure on laptop via websitemp4)
Here the interactive behaviors of the user include theviewing angle clipping and coloropacity transformationThefirst two items were operated on the Slave model which wasreconstructed on the client side While the user wants toobserve the final required image of high quality the userjust presses the submit button to submit the current behaviorrequest instruction to the server and waits for the final 2Dprojection image to be updated
Figure 6(a) presents the initial state of the Slave model(right) and 2D projection image (left) In Figure 6(b) mouseor hand gesture is used to operate the Slave model to
International Journal of Telemedicine and Applications 7
Table1Re
lationald
atas
torage
structureo
frequestbehaviorsfrom
clientand
resource
allocatio
non
thes
erver
Table
Field
Note
Basetable
DIC
OMs_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
grou
pof
SMIs
Torecord
theb
asicinform
ationof
SMIs
onthes
erverside
Storagep
ath
Thes
torage
path
ofSM
Ison
thes
erver
side
Pipelin
etable
pipelin
e_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
pipelin
efor
view
ingag
roup
ofSM
Isbetweenthec
lient
andserver
Basedon
basetableto
record
the
conn
ectedrelationshipbetweentheS
MIs
rend
eringin
mem
oryon
thes
ervera
ndremoteu
sero
nthec
lient
DIC
OMs_ID
FORE
IGNKE
Yof
basetableto
decla
rewhich
grou
pof
SMIsthec
lient
isview
ing
Client
Client
user
IPaddress
IPaddresso
fthe
committer
State
Whether
theS
MIsfin
ished
rend
eringin
mem
oryon
thes
erverYesallowthe
clienttonext
operation
Noprompt
for
waitin
g
Behavior
table
behavior_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofeach
behavior
requ
estevent
from
the
client
Basedon
pipelin
etable
torecord
operationrequ
estsfro
mthec
lient
user
andtheirp
rocesssta
te
pipelin
e_ID
FORE
IGNKE
Yof
pipelin
etable
todecla
rewhich
pipelin
eonview
ingand
processin
g
Behavior
type
Inclu
ding
view
inga
ngle
clippingor
coloropacity
transfo
rmation
Behavior-in
structio
n
From
thec
lienttotransfe
rthe
parametersinterm
softhe
correspo
nding
behavior
typerecordedin
theform
ofqu
antizationparameters
State
Pend
ingprocessin
gor
finish
ed
8 International Journal of Telemedicine and Applications
(a) Initial state (b) Viewing angle interaction
(c) Clipping interaction (d) Zooming and coloropacity interaction
Figure 6 Screenshot and operational demonstration
transform the viewing angle from Figure 6(a) to Figure 6(b)(red solid box) and then the submit button is pressed to sendthe behavior request instruction to the server to require anupdated projection image from the Master volume (yellowsolid box) In Figure 6(c) using virtual control panel (reddashed box) to operate Clipping plane after submitting thenew projection image was then updated (left) In Figure 6(d)the mouse wheel or hand gesture is used to zoom the Slavemodel and then the coloropacity theme is submitted via thedropdown list (bottom right corner) to guide remote volumerendering
4 Mode of Availability of the System
In this paper we designed aMaster-Slave dual-channel inter-action mode to improve a problem of remote visualizationinteraction of the volume reconstruction of SMIs whichusually have a huge data size and calculation cost Finallya visualization interaction platform has been built Particu-larly there is no special software and hardware requirementfor the clients and no special network environment isneeded
Tomeasure whether the effectiveness and performance toadapt to the mobile Internet application has characteristicsof lightweight three radiologists with more than 3 years ofexperience were invited to test the interactive experienceimaging quality and compatibility over diversified Internetdevices Moreover four junior students majoring in Biomed-ical Engineeringwere invited to test and quantize the networkload and response time
TestingData 3 groups of SMIs from the radiology departmentof Southwest Hospital of Chongqing include the following
(1) head and neck CT with 495 slices 512 lowast 512 2494MB(2) trunk CT with 609 slices 512 lowast 512 3065MB and (3)thorax andmandible CT with 500 slices 512lowast512 2516MBServer An IBM X3650M4 workstation (CPU 4 times E5-2603RAM 16GB OS Windows Server 2008) was connectedto the Internet via 10Mbps The 2D projection image wasset as the resolution of 512 lowast 512 with high quality JPEGcompression (level-8)
Client 4 Internet devices were located at the 3G or WIFImobile Internet with the server In Figure 7(a) there is aFounder E520 personal computer running on winXP withFirefox 46 web browser via 3G usb card In Figure 7(b) thereis an Acer E15 laptop running on win7 with chrome 43 webbrowser at WIFI In Figure 7(c) there is a HuaWei Honor6 smart phone running on EMUI 30 which is developedfrom Android 442 with the Opera 37 web browser at 3Gnet In Figure 7(d) there is an iPad mini running on IOS6with safari at WIFI All of them could be representative inthe mobile Internet Additionally each was limited to a 100-KBps wireless bandwidth by 360 security firewall for extremetest [33]
41 Interactive Experience Imaging Quality and Compatibil-ity To ensure the consistency of the context the testing data(1) were considered as the main object of reference Here thedemonstration and effect of the testing data (1) are shown inFigure 6
Figure 6 shows that this Slave model of the testing data(1) which was compressed to 104MB could be fit to thedemands of interactive navigation to accurately guide remotevolume rendering and shows that the projection image couldbe displayed to keep the original quality
International Journal of Telemedicine and Applications 9
(a) Personal computer (b) Laptop computer
(c) Smartphone (d) iPad
Figure 7 The system is accessed via four different devices and web browsers under extreme bandwidth
For further validation three radiologists were invited totest the platform with the similar behaviors in Figure 6 Thisjob was repeated to operate three groups of different testingdata through different web browsers and Internet devicesthat were all aforementioned Additionally a five-pointscale was designed to respond to the radiologistsrsquo requests(Table 2)
Table 2 presents a very positive evaluation from the statis-tical result of the imaging quality and interactive experienceand all three radiologists expressed that is an interestingway whether as a gesture-based or mouse-based interactionNo installations were required from the client and goodcompatibility was obtained (same result is shown in Table 3)However one radiologist thought the default Slave modelof the testing data (1) with a skeleton model of 104MB(Figure 3(d2)) was a little rough He suggested to use theskeleton model of 526MB or skin model of 913 KB (Figures3(c1) and 3(c2)) instead of the 104 model both of which arethe alternatives from the mutual interface for the client asdetailed in Section 231 Additionally all three radiologistsclaimed that the response time for final high quality image isalways about 1 second that is consistent with the test resultin Section 42 However they did accept the performancebecause they could directly operate the Slavemodel smoothlywithout any delay on client side thus a short wait for finalhigh quality image had little influence Moreover one of theradiologists specifically expressed the native application of3D rendering on desktop PC also needs 1 or 2 secondsrsquo delayfor final image this performance of the solution running onsmart phone is similar to the native application on PC and isvery good
42 Response Time and Network Load between the Clientand Server The response time can be defined as the timeinterval between the user sending a behavior request tothe system and web browser loading the response resultscompletely The network load can be defined as the amountof data transmission of each interactive behavior Four juniorstudents majoring in Biomedical Engineering were invitedto simultaneously test the platform for one hour During thetest all of the students respectively viewed and processed thetesting data (1) through the platform and performed similarbehaviors to those detailed in Figure 6This test was repeatedby each student with different web browsers and Internetdevices all of which were aforementioned Here we trainedthe students to record the real-time network traffic by the 360security firewall and the time interval by the script log Theresults are showed in Table 3
In Table 3 for each device the amount of data trans-mission and response time of the first loading are muchhigher than the behavior request and idle status because thefirst loading asks the client to download the necessary Slavemodel The response time of the first loading is lt12 s at 100-KBps client bandwidth which is similar to the total timespent on Slavemodel downloading In fact the response timeof first loading also includes the cost time of SMI loadingand rendering on the server side but the server monitorpresents the cost time as lt7 s because SMI loading and Slavemodel downloading are simultaneous and this cost could beneglected
After the first loading the network load of the behaviorrequest for final image is spent on request instructions sub-mission final projection image downloading and the serverprocessing status querying at a fixed frequency of 300ms
10 International Journal of Telemedicine and Applications
Table 2 Statistic results of the five-point scale concerning the interactive experience and imaging quality
Focus Subject Votes from five-point scaleVeryagree Agree Uncertainty Disagree Very
disagree
Interactive experienceMouse-basedGesture-based 998771
The operation of viewing and processing isaccurate
3998771 3
Simple and easy to use without training 2 1998771 3
A good user experience similar to nativeapplication
2 1998771 2 1
Imaging quality
Slavemodel could meet demands of navigation 2 12D image fromMaster volume could provide highquality reconstruction result via utilizing thedisplay capability of Internet terminals
3
Others Latency time for final high quality images 3
Table 3 Average response time and amount of data transmission from the different behaviors and devices at 100-KBps
Devices
First loading (averagevalue)
Behavior request for finalimage (average value)
Idle status ofbandwidthoccupation(KBps)
Frames PerSecond (FPS)Amount of
data trans (KB)Response time(seconds)
Amount ofdata trans (KB)
Response time(seconds)
PC (3G net) 108544 1123 4710 112 023 48ndash60Laptop (WIFI) 106496 1051 4608 089 031 48ndash60iPad mini (WIFI) 107941 1093 4403 095 035 50ndash60Smartphone (3G net) 107502 1072 3994 102 028 30ndash40
where the projection image downloading is the main partThe results present a complete amount of data transmission ofeach behavior request is close to 50KB but the correspondingresponse time for final image is approximately 1 s in 100KBpsThat is because besides the network load the rest of timeis spent on server processing and time interval of querying(300ms) The short delay could be acceptable for practicalapplications as detailed in Section 41
While in the idle status including the userrsquos interactivebehavior (operation) on the Slave model the bandwidthoccupation is almost negligible Moreover there are no sig-nificant differences among diversified devices and networksat 100-KBps client bandwidth
Additionally the FPS of Slave model rendering on theclient side is measured via statsminjs which is a librarypackaged from WebGL When the client rendering is com-pleted FPS could remain at the upper limit
5 Discussion and Comparison
Due to the nonprivate network environment and diversifiedpurposes application software on one device must competefor limited resources of the network and computing Thuslightweight which requires lower resources of the networkand computing and avoids unnecessary installation could bean important characteristic of mobile Internet applicationAdditionally to avoid compatibility problems caused by dif-ferent operating systems and client hardware characteristics
using HTML5 and related web technologies running onmajor web browsers may be a choice to adapt to cross-platform deployment and the characteristics of lightweight[21]
Therefore this article described research on a methodfor the Master-Slave dual-channel interaction mode whichis based on the HTML5 standard (also a wider web com-patibility standard of WebGL) to permit users viewing andprocessing million-megabyte-class SMIs via different clientdevices (without hardware or operating system constraints)over a pure web page (without installing special software) andin a diversified network environment
Compared with the traditional PACS mode of ldquoimagecompression transmission local reconstructionrdquo becauseour method does not transfer original SMIs to the clientit neither needs to wait for a long time to download norrequests special demands of the software environment onthe client device Moreover compared with the state ofthe art of local volume rendering based on WebGL andHTML5 technology [7ndash9] our methods have no specialdemands of client hardware for computation that may reducebattery consumption [9] and enhance the compatibility ofthe client devices our methods may be more fit for mobileInternet application Additionally the client does not accessthe original SMIs which could be much safer
Compared with popular solutions based on absolutesurface rendering technology [11 12 14] our method haslocally used this technique to only support light Slavemodel
International Journal of Telemedicine and Applications 11
for interactive navigation therefore the demands of thenetwork bandwidth and device rendering capability arealmost negligible in ourmethodMoreover the final requiredimage of high quality is a projection drawing from volumerendering technology indicating that our method could beconvenient for users to view the internal anatomy structureand complex organization relationship however surfacerendering technology could not support these characteristicsof volume
Compared with solutions based on absolute remote vol-ume rendering methods our method adopted the Master-Slave dual-channel interaction mode to strengthen the userexperience which is an entirely different approach from themode of ldquocontinuous request continuous responserdquo of studiesof [9] Although studies of [17ndash19] proposed several methodsto improve the mode to reduce the network load includingvideo-compressed transmission [17] tile-based transmission[18] and variable resolution transmission [19] the pressureof server load and real-time data stream increased We usedthe same SMI data (testing data (1)) the same quality offinal required image and the same server configurationand volume rendering algorithm in Section 4 to test themethods in [17ndash19] For quantitative test we formulated aninteractive trajectory with fixed 30 request points theoreti-cally assuming 30 rendering requests triggered per 1 secondthere should be 30 fps render effects On this basis [17]adopted 512 lowast 512 mpeg-4 compression for transmission[18] adopted 8 lowast 8 blocks to segment each 512 lowast 512 frameand transfer the blocks which are different from previousframe in the process of interaction and [19] adopted 64 lowast64 resolution transmission in the process of interaction but512 lowast 512 for end frame The test was repeated for 5 timesunder the condition of one single client monopolizing thebandwidth and server under ideal WIFI environment Theresults showed that the total amount of data transmissionfor each interaction test of [17ndash19] was 2980 KB 6036 KBand 3713 KB respectively the average response time of firstframe was 110 seconds 059 seconds and 052 seconds andthe average cumulative response time of end frame (30thframe which is the final required image) was 2131 seconds1761 seconds and 1601 seconds The response time containsthe factors of server continuous rendering response servercontinuous frame-encoding and so forth (more than 70server CPU utilization during interaction) but the networkfactors could be tentatively neglected under the ideal single-user WIFI environment Assuming that we can improvethe serverrsquos processing ability to completely eliminate theserver response delay (perfect 30 fps interaction effect) thetheoretical value of the bandwidth could be 2980 KBps6036 KBps and 3713 KBps respectively In summary theFPS value one of the most important parameters of the 3Dinteractive experience of the studies of [17ndash19]was correlatedreciprocally with the number of simultaneous tasks on theserver side (server task-pipelines) server processing abilityand network bandwidth and an obvious delay of interactionwas clearly presented By contrast all of the Internet deviceschosen to test our method could provide an FPS value at30ndash60 and the FPS valuewas not correlatedwith server tasksserver processing ability and network bandwidth but with
the device GPU (Graphics Processing Unit) Undoubtedlyalmost all GPUs of Internet devices in the market could meetthe demands of our light Slave model rendering Moreoverthe peak value of the bandwidth under the behavior requestfor final image was less than 50KBps and less than 1 KBpsat idle (including the operation of Slave model) thus bothwere far below those tests of methods of [17ndash19] Thusour method could be more suitable for a complex mobileInternet environment Moreover the complex compressiontransmission methods of [17ndash19] were built on customizedsoftware of plug-ins for the client which is hard to applyto pure web pages and falls short of wide compatibility Bycontrast that is our methodrsquos advantage
For the server load because the Master volume in ourmethod only responds to the final behavior request (instruc-tions) instead of the continuous request of the studies of[9 17ndash19] the test showed that server CPUutilizationwas lessthan 20 at idle (including the operation of Slavemodel) andless than 30 for responding to behavior request thus it wasfar below of those tests ofmethods of [17ndash19] with nearly a fullload during all the operation time under the same conditionsof SMI data quality of final required image server config-uration and volume rendering algorithm Theoretically ourplatform can respond to more independent tasks (pipelines)at the same time Table 2 shows that while the server wasrunning four independent tasks simultaneously the responsetime of our method remained at approximately 1 s in the 100-KBps client bandwidth much better than [17ndash19] which waswith more than 16 secondsrsquo delay of end frame in the singletask status In fact because user viewing and processing ofthe 3D medical volume occurred via interaction on the Slavemodel the delay experience wasmuch fewer than that in [17ndash19] as detailed in Section 41That indicates ourmethod couldhave some more advantages including much less networkload and server load to serve much more independent tasksand users
In addition [17ndash19] did not propose an effective inter-active method to clip the 3D volume what our method hasaccomplished
6 Conclusion
This paper proposed aMaster-Slave dual-channel interactionmode which absorbed the advantages of the surface ren-dering of WebGL and remote volume rendering to builda visualization interaction platform to rapidly view andprocess 3Dmedical volume reconstruction in a pure webpageenvironment This platform could be used in a diversifiednetwork environment (100-KBps bandwidth for extreme)and diversified client devices (no special hardware and oper-ating system constraints) without any special installation butwith a good user experience similar to native applicationThese features could serve authorized users to convenientlyaccess SMIs and present 3D visualization anywhere andcould be a technological basis of online communicationamong doctors and patients [34] Or at least it could helpclinicians to arbitrarily acquire 3D image according to theirclinical needs instead of the static pictures that radiologistsdrafted For example clinicians could conveniently observe
12 International Journal of Telemedicine and Applications
the complicated relationship of skull base bone and bloodvessel via this platform (Figure 6(d)) in the past this pieceof information was provided by a few static images generatedfrom radiologistrsquos understanding
Additionally this solution is designed to be followed asanHTML5 standard interface indicating any authorized usercould rapidly access the application via one URL hyperlinkand the platform could be easily integrated into othertelemedical systems seamlessly as a third-party applicationfor example the Supplementary Materials (2 Demonstrationof the operation procedure on laptop linked by emailmp4and 3 Demonstration of the operation procedure on iPadvia websitemp4) These features may play an assistant roleto improve work patterns and extend the range of applicationof traditional medical image systems For example it can beapplied to regional health telemedicine andmedical imagingeducation at a low cost because there is no special cost ofdeploy and upgrade of software network and hardware onthe client side And more importantly a teacher can vividlyshow the students the image characteristics of the diseaseanywhere a clinician can intuitively introduce disease to thepatients anywhere and ask expertsrsquo help anywhere That maypromote the development of mobile medical technology to acertain degree
Conflicts of Interest
There are no conflicts of interest declared
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China (no 81171866) the National Key BasicResearch Program of China (no 2014CB541602) and theSocial Livelihood Science andTechnology Innovation SpecialProject of CSTC (no cstc2015shmszx120002)
References
[1] Q Zhang M Alexander and L Ryner ldquoSynchronized 2D3Doptical mapping for interactive exploration and real-time visu-alization of multi-function neurological imagesrdquo ComputerizedMedical Imaging and Graphics vol 37 no 7-8 pp 552ndash567 2013
[2] S Zimeras and L G Gortzis ldquoInteractive tele-radiological seg-mentation systems for treatment and diagnosisrdquo InternationalJournal of Telemedicine and Applications vol 2012 Article ID713739 p 15 2012
[3] S E Mahmoudi A Akhondi-Asl R Rahmani et al ldquoWeb-based interactive 2D3D medical image processing andvisualization softwarerdquo Computer Methods and Programs inBiomedicine vol 98 no 2 pp 172ndash182 2010
[4] L Faggioni E Neri I Bargellini et al ldquoIPad-based primary2D reading of CT angiography examinations of patients withsuspected acute gastrointestinal bleeding preliminary experi-ence 1Lrdquo British Journal of Radiology vol 88 no 1047 ArticleID 20140477 2015
[5] J B Park H J Choi J H Lee and B S Kang ldquoAn assessment ofthe iPad 2 as a CT teleradiology tool using brain CT with sub-tle intracranial hemorrhage under conventional illuminationrdquoJournal of Digital Imaging vol 26 no 4 pp 683ndash690 2013
[6] P T Johnson S L Zimmerman D Heath et al ldquoThe iPad asa mobile device for CT display and interpretation diagnosticaccuracy for identification of pulmonary embolismrdquo EmergencyRadiology vol 19 no 4 pp 323ndash327 2012
[7] J M Noguera and J R Jimenez ldquoVisualization of very large 3Dvolumes on mobile devices and WebGLrdquo in Proceedings of the20th International Conference in Central Europe on ComputerGraphics Visualization and Computer Vision WSCG 2012 pp105ndash112 June 2012
[8] J Congote A Segura L Kabongo et al ldquoInteractive visu-alization of volumetric data with WebGL in real-timerdquo inProceedings of the 16th International Conference on 3D WebTechnology (Web3D rsquo11) pp 137ndash145 June 2011
[9] A Schiewe M Anstoots and J Kruger ldquoState of the art inmobile volume rendering on iOS devicesrdquo in Proceedings of theEurographics Conference on Visualization (EuroVis rsquo15) 2015
[10] A Evans M Romeo A Bahrehmand et al ldquo3D graphics on theweb a surveyrdquo Computers amp Graphics vol 41 no 1 pp 43ndash612014
[11] M Callieri R M Andrei M D Benedetto et al ldquoVisualizationmethods formolecular studies on theweb platformrdquo inProceed-ings of the 15th international conference on web 3D technology(Web3D rsquo10) pp 117ndash126 July 2010
[12] J Jimenez A M Lopez J Cruz et al ldquoA Web platform forthe interactive visualization and analysis of the 3D fractaldimension of MRI datardquo Journal of Biomedical Informatics vol51 pp 176ndash190 2014
[13] WebGL httpswwwkhronosorgwebgl[14] C R Butson G Tamm S Jain T Fogal and J Kruger
ldquoEvaluation of interactive visualization on mobile computingplatforms for selection of deep brain stimulation parametersrdquoIEEE Transactions on Visualization and Computer Graphics vol19 no 1 pp 108ndash117 2013
[15] M Levoy ldquoDisplay of surfaces from volume datardquo IEEE Com-puter Graphics and Applications vol 8 no 3 pp 29ndash37 1988
[16] E Kotter T Baumann D Jager and M Langer ldquoTechnologiesfor image distribution in hospitalsrdquo European Radiology vol 16no 6 pp 1270ndash1279 2006
[17] S Park W Kim and I Ihm ldquoMobile collaborative medical dis-play systemrdquo Computer Methods and Programs in Biomedicinevol 89 no 3 pp 248ndash260 2008
[18] J R Mitchell P Sharma J Modi et al ldquoA smartphone client-server teleradiology system for primary diagnosis of acutestrokerdquo Journal of Medical Internet Research vol 13 no 2 p e312011
[19] T Hachaj ldquoReal time exploration and management of largemedical volumetric datasets on small mobile devicesmdasheval-uation of remote volume rendering approachrdquo InternationalJournal of Information Management vol 34 no 3 pp 336ndash3432014
[20] P Quax J Liesenborgs A Barzan et al ldquoRemote renderingsolutions using web technologiesrdquoMultimedia Tools and Appli-cations vol 75 no 8 pp 4383ndash4410 2016
[21] D Preuveneers Y Berbers and W Joosen ldquoThe future ofmobile E-health application development exploring HTML5for context-aware diabetes monitoringrdquo Procedia ComputerScience vol 21 pp 351ndash359 2013
[22] J M Noguera J J Jimenez and M C Osuna-Perez ldquoDevel-opment and evaluation of a 3D mobile application for learningmanual therapy in the physiotherapy laboratoryrdquo Computersand Education vol 69 pp 96ndash108 2013
International Journal of Telemedicine and Applications 13
[23] Y S Kang ldquoVolume rendering overviewrdquo inGPU Programmingand Cg Language Primer chapter 14 pp 154-155 1st edition2009
[24] Visualization Toolkit (VTK) httpwwwvtkorg[25] vtkPiecewiseFunction Class Reference of VTK httpwwwvtk
orgdocnightlyhtmlclassvtkPiecewiseFunctionhtml[26] vtkColorTransferFunction Class Reference of VTK http
wwwvtkorgdocnightlyhtmlannotatedhtml[27] vtkPlane Class Reference of VTK httpwwwvtkorgdoc
nightlyhtmlclassvtkPlanehtml[28] L Ma D-X Zhao and Z-Z Yang ldquoA software tool for
visualization of molecular face (VMF) by improving marchingcubes algorithmrdquo Computational and Theoretical Chemistryvol 1028 pp 34ndash45 2014
[29] G F Wang J Liu and W Liu ldquoDIB display and window trans-formation technology for DICOM medical imagesrdquo ChineseJournal of Medical Device vol 18 no 8 pp 1ndash5 2005
[30] L Qiao X Chen L X Yang et al ldquoDevelopment of medi-cal image three-dimensional-visualization platform based onhome network and pure web conditionsrdquo Beijing BiomedicalEngineering vol 34 no 3 pp 229-233286 2015
[31] M Ying and J Miller ldquoRefactoring legacy AJAX applications toimprove the efficiency of the data exchange componentrdquo Journalof Systems and Software vol 86 no 1 pp 72ndash88 2013
[32] HTML5 httpwwww3orgTR2012CR-html5-20121217[33] QIHU 360 SOFTWARE CO LIMITED 360 Internet Security
httpwww360safecomInternet-securityhtml[34] L Qiao Y Li X Chen et al ldquoMedical high-resolution image
sharing and electronic whiteboard system a pure-web-basedsystem for accessing and discussing lossless original images intelemedicinerdquoComputerMethods and Programs in Biomedicinevol 121 no 2 pp 77ndash91 2015
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of Telemedicine and Applications 3
Interactivebehaviors
Sectional medical images (SMIs) amp Slave model
Service
N
M
Any device amp OS amp browser
e screen eect
2D imageMaster volume
Listener
Projection
Slave model
Internet
Client
Volume rendering
Behaviorrequest
Request Mgmt
Loading in memory
1
2
3
85
86
middot middot middot
middot middot middot
middot middot middot
12
3
85
86
12
3
8586
middot middot middot
①
④
⑤
③
②
Figure 1 Architecture of the pure web-based solution via theMaster-Slave dual-channel interaction mode
Here the client could be a pure web browser running ondiversified devices supporting the WebGL technique Thatmeans there is no installation for the client
22 Rendering of the Master Volume We adopted the ray-casting algorithm [22] to reconstruct the Master volume ofSMIsThe algorithm is a direct volume renderingmethod thathas been widely used for radiodiagnosis but requires a highcomputing performance of the device First we transformedthe SMIs into volume voxels in 3D space (Figure 2(a)) andthen simulated rays to cast each voxel with a customizedcolor and opacity [23] into 2D space (Figure 2(b)) Thismethod could be used to describe a complex organizationrelationship Furthermore it could display internal anatomystructure via removing (clipping) part of the voxels thatsimulate a userrsquos behavior of dissection (Figure 2(c))
Here we directly used the VTK (Visualization Toolkit)library [24] to realize the ray-casting algorithm For cus-tomized color and opacity several frequently used schemescould be deployed on the server side in advance a dropdownlist was designed on the user interface to correspond to eachscheme for remote volume rendering [25 26] For clippingVTK provides a class of vtkPlane to build a virtual Clippingplane to guide the voxel removal [27] It means the clippingjob only needs the specific location of the virtual Clippingplane in the 3D coordinate systemThe specific location couldbe described by the following two parameters in the computergraphics method normal vector from the origin point to theplane and the coordinate position of a point on the Clippingplane both from the client submission
23 Generation and Interaction of the Slave Model The Slavemodel was pregenerated on the server side and was designedto be a navigational tool to interact with theWeb page Whileclients want to view and process theMaster volume the Slavemodel could be downloaded to the client and reconstructedon the client side
231 Generation and Slimming of the SlaveModel To interactwith mobile Internet the data size of the model should beas small as possible (much slimmer) while the clarity ofthe contour is acceptable However using the traditionalcompression-decompression algorithm may ask the client toinstall additional software Hence the steps in the generationof a smaller Slavemodel can be designed as follows (Figure 3)Here we used a group of SMIs of the head and neck CT asan example and the resolution of voxels is 512 times 512 times 495249MB (testing data (1) in Section 4)
Step 1 (resampling the original SMIs for first slimming) Loadthe corresponding SMIs into memory shrink each sectionalimage to a standard size of 32 lowast 32 via the resampling of the2D space (Figures 3(a) and 3(b))
Step 2 (building the 3D geometrical contour model) Amarching cubes algorithm [28] one of the most commonmethods of surface rendering was used to extract theinterested surface contour of the anatomical representationfrom resampling SMIs into a 3D geometrical contour modelHere according to the feature of the CT grayscale [29] weset a particular contour value at minus800 or 200 that couldseparate the skin or skeleton of the body from the resampling
4 International Journal of Telemedicine and Applications
(a) Voxels in volume data (b) Pseudo color amp opacity for each voxel (c) Clipping
Figure 2 Schematic diagram of the direct volume rendering method
Size YSize X
Size ZSize Z
Size XSize Y
Y Y
X X
Z Z
(a) Original SMIs
Step 1 Resampling the original SMIs for rst compression
Shrink each slice
(b) Resampling amp strech
(c1) Contour value minus800
(c2) Contour value 200Size 526MBStep 2 Building
3D contour model
(d1) Polygons 50 Polygons 10
(d2) Polygons 50 Polygons 10
Step 3 Reduce the triangles
Step 4Saved asvtk le
(eg 512 times 512 times 495 249 MB) 311MBSize 104MBSize
333MBSize 913MBSize674MBSize
2 times 32 times 495)(eg 3
Figure 3 Schematic diagram of the generation and compression of the Slavemodel
SMIs respectively to acquire a skin model or skeleton model(Figures 3(c1) and 3(c2)) In this case the data size of the skinmodel and skeleton model could be reduced to 674MB and526MB respectively
Step 3 (reducing the total number of polygons for secondslimming) The 3D geometrical contour model from Step 2is constitutive of polygons and we could reduce the totalnumber of polygons to further reduce the data size In thiscase the number was reduced to 50 and 10 and thedata sizes were reduced to approximately 3MB and 1MBrespectively (Figures 3(d1) and 3(d2))
Step 4 (saved as vtk model file) The slimmed model wouldbe saved in a file of the vtk format which is a storage formatof the geometrical model supported by WebGL and thenstored in the same directory with original SMIs
The result of Figures 3(d1) and 3(d2) preliminarily showedthat the Slave model whose data size is approximately 1MBstill could be fit for interactive navigation (more experimentsdetailed in Section 41) Moreover an additional mutualinterface of the client was designed as a remote parametersetting of the contour value and the total number of polygonsto permit users to customize the Slavemodel as desired
232 Design and Quantization of the Interactive Behaviors inthe Slave Model For a friendly interactive experience on theclient side a simple style of interactive behaviors of the clientand corresponding quantitative methods should be designed
to guide remote volume rendering including the viewingangle request and clipping request
For the Viewing Angle Request The interaction of viewingangle includes rotation flipping zooming and panning [30]We defined the interactive 3D scene including only one Slavemodel and one camera and customized an interactive modeof the camera in VR technology to meet our requirementFigure 4(a) shows the principle of the interaction of theviewing angle between the camera and Slave model Welimited the lens of the camera to always be taken on point1198751 (0 0 0) which was the origin of the world coordinatesystem The interactive behaviors of the viewing angle werecarried out by only changing the camera position exceptpanning For panning it was carried out by changing theSlavemodel position see examples in Figure 4(b)
According to the definition while sending requestinstructions to the server for remote volume rendering theinstructions only include three quantization parameters fromthe client to accurately describe the interactive behavior ofthe viewing angle request to the Slavemodel The three quan-tization parameters include the following attribute view updirection attribute position of camera and attribute positionof the center point of the Slavemodel
Here attribute view up direction of camera indicates thedirection of the top of the camera For example (0 minus1 0)a value of the View Up direction indicates the camera isperpendicular to the XZ plane in the world coordinatesystem and the top of the camera is in the opposite direction
International Journal of Telemedicine and Applications 5
(a) Principle of interaction between camera and Slavemodel
(i) Original camera view
(iii) Zoom via shorteningthe distance from CaA to P1
(iv) Flip from CaA to CaB (v) Panning via movingthe center point of Slavemodel from P1 to P2
(ii) Rotate around axis CaA-P1 (Y-axis)
(b) Example of interactions of viewing angle request
Figure 4 Principle of the interaction of the viewing angle between the camera and Slavemodel and the corresponding example
along the 119885-axis (see the relationship between camera CaAand Slave model in Figure 4(a) and corresponding example(i) original camera view in Figure 4(b))
In fact WebGL has provided some interaction methodsof the moving camera to view 3D objects via mouse- andgesture-based interactive behaviors which were packagedinto a library of TrackballControlsjs for a web page scriptWe could directly utilize the js library to acquire the threequantization parameters However we need to rewrite thejs library to restrict and customize the interactive modeof the camera that we have defined to ensure the consis-tency of the interactive rule between the Master and Slaveside
For the Clipping Request For arbitrarily clipping and viewingthe internal anatomy structure of theMaster volume a greenrectangle object named the Clipping plane was added in theabove defined interactive 3D scene to allow users to mimicthe clipping behavior in the Slavemodel see Figures 5(a) and5(b) Here a virtual control panel was designed separately tosupport the interactive behaviors of users using the Clippingplane (Figure 5(c))
During the initialization phase the Clipping plane coversthe 119883119884 plane in the world coordinate system (Figure 5(a))Its specific location could be accurately described by thefollowing two quantization parameters in the computergraphics method normal vector from the origin point tothe Clipping plane and the coordinate position of a pointon the Clipping plane Here its default normal vector was0 0 1 one point 119875 on the plane was on the position(0 0 0)
For intuitive operation of the user experience the interac-tive behaviors regarding the Clipping plane include rotationand panning and both can be operated by a virtual controlpanel However for the accurate description of the requestinstructions for remote volume rendering the two quantiza-tion parameters should be used
Therefore formula (1) was deduced to calculate thenormal vector from the angles of rotation
119899119900119903119898119886119897 V119890119888119905119900119903 = sin (120579119910) minus sin (120579119909)
sdot cos (120579119910) cos (120579119909) sdot cos (120579119910) (1)
Here 120579119909 120579119910 are the final rotation angles that rotate theClipping plane clockwise around the 119883-axis and 119884-axisrespectively
For example in Figure 5(b) a user rotated the Clippingplane 30 degrees clockwise around the 119884-axis and minus15degrees clockwise around the 119883-axis and then panned theClipping plane 95 pixels in the positive direction along the119885-axis While submitting for remote volume rendering 120579119909 =minus15 120579119910 = 30 the normal vector was 05 02241 08365according to formula (1) Meanwhile according to the char-acteristic of monolithic translation the one point 119875 on theClipping plane was moved to position (0 0 95) That wouldbe the two quantization parameters of request instructions forclipping
24 Remote Interaction between the Server and Client Theinteractive behaviors of the Slavemodel include viewing anglerequest and clipping request Moreover coloropacity request isoperated by the dropdown list of the preset schemes (detailedin Section 22) All of these requests would be submitted tothe server in the form of quantization parameters which arenamed as request instructions
However in addition to responding to the clientrsquos instruc-tions the server has to find an efficient way to manageand allocate system resources to enhance the robustness ofthe high frequency of interaction among multiple users forexample to reduce the probability of deadlock cumulativedelay and task-pipeline crossing
In our system we used relational database schema tomanage the request instructions from the client and to
6 International Journal of Telemedicine and Applications
Interactiveeect
Clippingplane
XY
Z
PX
Y
Z
(a) Initial state of Clipping plane
PX
Y
Z
(i) Rotated 30 degrees (ii) Further rotated minus15 degreesaround the X-axis
(ii) Further panning 95
P
XY
Z
P
XY
Z
Y-axisaround the Z-axisalong the
(b) Instances of interaction of Clipping plane (c) Control panelof Clipping plane
Figure 5 Design of the interaction of clipping to the Slavemodel and corresponding example
allocate the system resources for guiding 3D rendering Thecore tables and their data storage structure are designed inTable 1
According to Table 1 if a user chooses a group of SMIsfrom a list derived from the base table the server wouldquery the pipeline table whether the SMIs is serving theclient If No the SMIs would be loaded in memory anda task-pipeline between the user and SMIs would be builtThereafter the state of the pipeline table would be changed toldquoYesrdquoWhile being Yes the client would be permitted tomakethe next interactive request After that for each request fromthe client the servermust query the state of the behavior tablewhether the pipeline is ldquopendingrdquo or ldquoprocessingrdquo If No theserver would accept the request with a record of a ldquopendingrdquostate Additionally the listener on the server would real-timearrange these pending requests to the server rendering inqueue While a request is being rendered the correspondingstate of behavior table would be changed as ldquoprocessingrdquo toprevent malignant submission While the job is finished thestate would be changed to ldquofinishedrdquo and a 2D projectionimage would be generated and saved in the specified HTTPsharing path
The client adopted AJAX (Asynchronous JavaScript andXML) [31] to dynamically query the state of behavior tablefrom the server at a fixed frequency of 300ms While render-ing was finished the client would download and update the2D projection image via the HTML + js web page technique
3 System Description
31 System Architecture While Figure 1 depicts the Master-Slave dual-channel interaction mode it also presents theoverall architecture of the final platform Here we adoptedthe browserserver architecture for which the communi-cation with each other was via the HTTP protocol Theserver side consists of the Listener amp Request ManagementServer (in zone B) and Rendering Server (in zone C) Theformer accepts and manages the requests from the client
and guides the Rendering Server to an orderly responseThe latter renders the SMIs and generates a 2D projec-tion image for the client downloading via HTTP At thesame time the client side was designed by the standardof HTML5 [32] We used ltimggt an HTML image tagto display the 2D projection image in the form of pseudo3D and used AJAX + js technology to realize flicker-freepage updates and client RIA (rich Internet application)over diversified web browsers without installation More-over HTML5 includes the WebGL standard and the Slavemodel could run on any HTML5-supporting web browserswithout any web plug-ins or special software tools Thusthe whole platform system could run under mature networktechnologies
32 System Front-End System front-end is anHTML5-basedpure website that provides entrance to SMI access and 3Dpresentation The users require a username and password toenter the system or any experts could be invited to enter aspecific task-pipeline with one URL hyperlink from a helpseeker In brief using an HTML5-supporting web browserincluding the major versions of Firefox chrome Operasafari IE 10+ the 3D volume can be viewed and processedas in Figure 6 and the Supplementary Materials availableonline at httpsdoiorg10115520174074137 (1 Demon-stration of the operation procedure on laptop via websitemp4)
Here the interactive behaviors of the user include theviewing angle clipping and coloropacity transformationThefirst two items were operated on the Slave model which wasreconstructed on the client side While the user wants toobserve the final required image of high quality the userjust presses the submit button to submit the current behaviorrequest instruction to the server and waits for the final 2Dprojection image to be updated
Figure 6(a) presents the initial state of the Slave model(right) and 2D projection image (left) In Figure 6(b) mouseor hand gesture is used to operate the Slave model to
International Journal of Telemedicine and Applications 7
Table1Re
lationald
atas
torage
structureo
frequestbehaviorsfrom
clientand
resource
allocatio
non
thes
erver
Table
Field
Note
Basetable
DIC
OMs_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
grou
pof
SMIs
Torecord
theb
asicinform
ationof
SMIs
onthes
erverside
Storagep
ath
Thes
torage
path
ofSM
Ison
thes
erver
side
Pipelin
etable
pipelin
e_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
pipelin
efor
view
ingag
roup
ofSM
Isbetweenthec
lient
andserver
Basedon
basetableto
record
the
conn
ectedrelationshipbetweentheS
MIs
rend
eringin
mem
oryon
thes
ervera
ndremoteu
sero
nthec
lient
DIC
OMs_ID
FORE
IGNKE
Yof
basetableto
decla
rewhich
grou
pof
SMIsthec
lient
isview
ing
Client
Client
user
IPaddress
IPaddresso
fthe
committer
State
Whether
theS
MIsfin
ished
rend
eringin
mem
oryon
thes
erverYesallowthe
clienttonext
operation
Noprompt
for
waitin
g
Behavior
table
behavior_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofeach
behavior
requ
estevent
from
the
client
Basedon
pipelin
etable
torecord
operationrequ
estsfro
mthec
lient
user
andtheirp
rocesssta
te
pipelin
e_ID
FORE
IGNKE
Yof
pipelin
etable
todecla
rewhich
pipelin
eonview
ingand
processin
g
Behavior
type
Inclu
ding
view
inga
ngle
clippingor
coloropacity
transfo
rmation
Behavior-in
structio
n
From
thec
lienttotransfe
rthe
parametersinterm
softhe
correspo
nding
behavior
typerecordedin
theform
ofqu
antizationparameters
State
Pend
ingprocessin
gor
finish
ed
8 International Journal of Telemedicine and Applications
(a) Initial state (b) Viewing angle interaction
(c) Clipping interaction (d) Zooming and coloropacity interaction
Figure 6 Screenshot and operational demonstration
transform the viewing angle from Figure 6(a) to Figure 6(b)(red solid box) and then the submit button is pressed to sendthe behavior request instruction to the server to require anupdated projection image from the Master volume (yellowsolid box) In Figure 6(c) using virtual control panel (reddashed box) to operate Clipping plane after submitting thenew projection image was then updated (left) In Figure 6(d)the mouse wheel or hand gesture is used to zoom the Slavemodel and then the coloropacity theme is submitted via thedropdown list (bottom right corner) to guide remote volumerendering
4 Mode of Availability of the System
In this paper we designed aMaster-Slave dual-channel inter-action mode to improve a problem of remote visualizationinteraction of the volume reconstruction of SMIs whichusually have a huge data size and calculation cost Finallya visualization interaction platform has been built Particu-larly there is no special software and hardware requirementfor the clients and no special network environment isneeded
Tomeasure whether the effectiveness and performance toadapt to the mobile Internet application has characteristicsof lightweight three radiologists with more than 3 years ofexperience were invited to test the interactive experienceimaging quality and compatibility over diversified Internetdevices Moreover four junior students majoring in Biomed-ical Engineeringwere invited to test and quantize the networkload and response time
TestingData 3 groups of SMIs from the radiology departmentof Southwest Hospital of Chongqing include the following
(1) head and neck CT with 495 slices 512 lowast 512 2494MB(2) trunk CT with 609 slices 512 lowast 512 3065MB and (3)thorax andmandible CT with 500 slices 512lowast512 2516MBServer An IBM X3650M4 workstation (CPU 4 times E5-2603RAM 16GB OS Windows Server 2008) was connectedto the Internet via 10Mbps The 2D projection image wasset as the resolution of 512 lowast 512 with high quality JPEGcompression (level-8)
Client 4 Internet devices were located at the 3G or WIFImobile Internet with the server In Figure 7(a) there is aFounder E520 personal computer running on winXP withFirefox 46 web browser via 3G usb card In Figure 7(b) thereis an Acer E15 laptop running on win7 with chrome 43 webbrowser at WIFI In Figure 7(c) there is a HuaWei Honor6 smart phone running on EMUI 30 which is developedfrom Android 442 with the Opera 37 web browser at 3Gnet In Figure 7(d) there is an iPad mini running on IOS6with safari at WIFI All of them could be representative inthe mobile Internet Additionally each was limited to a 100-KBps wireless bandwidth by 360 security firewall for extremetest [33]
41 Interactive Experience Imaging Quality and Compatibil-ity To ensure the consistency of the context the testing data(1) were considered as the main object of reference Here thedemonstration and effect of the testing data (1) are shown inFigure 6
Figure 6 shows that this Slave model of the testing data(1) which was compressed to 104MB could be fit to thedemands of interactive navigation to accurately guide remotevolume rendering and shows that the projection image couldbe displayed to keep the original quality
International Journal of Telemedicine and Applications 9
(a) Personal computer (b) Laptop computer
(c) Smartphone (d) iPad
Figure 7 The system is accessed via four different devices and web browsers under extreme bandwidth
For further validation three radiologists were invited totest the platform with the similar behaviors in Figure 6 Thisjob was repeated to operate three groups of different testingdata through different web browsers and Internet devicesthat were all aforementioned Additionally a five-pointscale was designed to respond to the radiologistsrsquo requests(Table 2)
Table 2 presents a very positive evaluation from the statis-tical result of the imaging quality and interactive experienceand all three radiologists expressed that is an interestingway whether as a gesture-based or mouse-based interactionNo installations were required from the client and goodcompatibility was obtained (same result is shown in Table 3)However one radiologist thought the default Slave modelof the testing data (1) with a skeleton model of 104MB(Figure 3(d2)) was a little rough He suggested to use theskeleton model of 526MB or skin model of 913 KB (Figures3(c1) and 3(c2)) instead of the 104 model both of which arethe alternatives from the mutual interface for the client asdetailed in Section 231 Additionally all three radiologistsclaimed that the response time for final high quality image isalways about 1 second that is consistent with the test resultin Section 42 However they did accept the performancebecause they could directly operate the Slavemodel smoothlywithout any delay on client side thus a short wait for finalhigh quality image had little influence Moreover one of theradiologists specifically expressed the native application of3D rendering on desktop PC also needs 1 or 2 secondsrsquo delayfor final image this performance of the solution running onsmart phone is similar to the native application on PC and isvery good
42 Response Time and Network Load between the Clientand Server The response time can be defined as the timeinterval between the user sending a behavior request tothe system and web browser loading the response resultscompletely The network load can be defined as the amountof data transmission of each interactive behavior Four juniorstudents majoring in Biomedical Engineering were invitedto simultaneously test the platform for one hour During thetest all of the students respectively viewed and processed thetesting data (1) through the platform and performed similarbehaviors to those detailed in Figure 6This test was repeatedby each student with different web browsers and Internetdevices all of which were aforementioned Here we trainedthe students to record the real-time network traffic by the 360security firewall and the time interval by the script log Theresults are showed in Table 3
In Table 3 for each device the amount of data trans-mission and response time of the first loading are muchhigher than the behavior request and idle status because thefirst loading asks the client to download the necessary Slavemodel The response time of the first loading is lt12 s at 100-KBps client bandwidth which is similar to the total timespent on Slavemodel downloading In fact the response timeof first loading also includes the cost time of SMI loadingand rendering on the server side but the server monitorpresents the cost time as lt7 s because SMI loading and Slavemodel downloading are simultaneous and this cost could beneglected
After the first loading the network load of the behaviorrequest for final image is spent on request instructions sub-mission final projection image downloading and the serverprocessing status querying at a fixed frequency of 300ms
10 International Journal of Telemedicine and Applications
Table 2 Statistic results of the five-point scale concerning the interactive experience and imaging quality
Focus Subject Votes from five-point scaleVeryagree Agree Uncertainty Disagree Very
disagree
Interactive experienceMouse-basedGesture-based 998771
The operation of viewing and processing isaccurate
3998771 3
Simple and easy to use without training 2 1998771 3
A good user experience similar to nativeapplication
2 1998771 2 1
Imaging quality
Slavemodel could meet demands of navigation 2 12D image fromMaster volume could provide highquality reconstruction result via utilizing thedisplay capability of Internet terminals
3
Others Latency time for final high quality images 3
Table 3 Average response time and amount of data transmission from the different behaviors and devices at 100-KBps
Devices
First loading (averagevalue)
Behavior request for finalimage (average value)
Idle status ofbandwidthoccupation(KBps)
Frames PerSecond (FPS)Amount of
data trans (KB)Response time(seconds)
Amount ofdata trans (KB)
Response time(seconds)
PC (3G net) 108544 1123 4710 112 023 48ndash60Laptop (WIFI) 106496 1051 4608 089 031 48ndash60iPad mini (WIFI) 107941 1093 4403 095 035 50ndash60Smartphone (3G net) 107502 1072 3994 102 028 30ndash40
where the projection image downloading is the main partThe results present a complete amount of data transmission ofeach behavior request is close to 50KB but the correspondingresponse time for final image is approximately 1 s in 100KBpsThat is because besides the network load the rest of timeis spent on server processing and time interval of querying(300ms) The short delay could be acceptable for practicalapplications as detailed in Section 41
While in the idle status including the userrsquos interactivebehavior (operation) on the Slave model the bandwidthoccupation is almost negligible Moreover there are no sig-nificant differences among diversified devices and networksat 100-KBps client bandwidth
Additionally the FPS of Slave model rendering on theclient side is measured via statsminjs which is a librarypackaged from WebGL When the client rendering is com-pleted FPS could remain at the upper limit
5 Discussion and Comparison
Due to the nonprivate network environment and diversifiedpurposes application software on one device must competefor limited resources of the network and computing Thuslightweight which requires lower resources of the networkand computing and avoids unnecessary installation could bean important characteristic of mobile Internet applicationAdditionally to avoid compatibility problems caused by dif-ferent operating systems and client hardware characteristics
using HTML5 and related web technologies running onmajor web browsers may be a choice to adapt to cross-platform deployment and the characteristics of lightweight[21]
Therefore this article described research on a methodfor the Master-Slave dual-channel interaction mode whichis based on the HTML5 standard (also a wider web com-patibility standard of WebGL) to permit users viewing andprocessing million-megabyte-class SMIs via different clientdevices (without hardware or operating system constraints)over a pure web page (without installing special software) andin a diversified network environment
Compared with the traditional PACS mode of ldquoimagecompression transmission local reconstructionrdquo becauseour method does not transfer original SMIs to the clientit neither needs to wait for a long time to download norrequests special demands of the software environment onthe client device Moreover compared with the state ofthe art of local volume rendering based on WebGL andHTML5 technology [7ndash9] our methods have no specialdemands of client hardware for computation that may reducebattery consumption [9] and enhance the compatibility ofthe client devices our methods may be more fit for mobileInternet application Additionally the client does not accessthe original SMIs which could be much safer
Compared with popular solutions based on absolutesurface rendering technology [11 12 14] our method haslocally used this technique to only support light Slavemodel
International Journal of Telemedicine and Applications 11
for interactive navigation therefore the demands of thenetwork bandwidth and device rendering capability arealmost negligible in ourmethodMoreover the final requiredimage of high quality is a projection drawing from volumerendering technology indicating that our method could beconvenient for users to view the internal anatomy structureand complex organization relationship however surfacerendering technology could not support these characteristicsof volume
Compared with solutions based on absolute remote vol-ume rendering methods our method adopted the Master-Slave dual-channel interaction mode to strengthen the userexperience which is an entirely different approach from themode of ldquocontinuous request continuous responserdquo of studiesof [9] Although studies of [17ndash19] proposed several methodsto improve the mode to reduce the network load includingvideo-compressed transmission [17] tile-based transmission[18] and variable resolution transmission [19] the pressureof server load and real-time data stream increased We usedthe same SMI data (testing data (1)) the same quality offinal required image and the same server configurationand volume rendering algorithm in Section 4 to test themethods in [17ndash19] For quantitative test we formulated aninteractive trajectory with fixed 30 request points theoreti-cally assuming 30 rendering requests triggered per 1 secondthere should be 30 fps render effects On this basis [17]adopted 512 lowast 512 mpeg-4 compression for transmission[18] adopted 8 lowast 8 blocks to segment each 512 lowast 512 frameand transfer the blocks which are different from previousframe in the process of interaction and [19] adopted 64 lowast64 resolution transmission in the process of interaction but512 lowast 512 for end frame The test was repeated for 5 timesunder the condition of one single client monopolizing thebandwidth and server under ideal WIFI environment Theresults showed that the total amount of data transmissionfor each interaction test of [17ndash19] was 2980 KB 6036 KBand 3713 KB respectively the average response time of firstframe was 110 seconds 059 seconds and 052 seconds andthe average cumulative response time of end frame (30thframe which is the final required image) was 2131 seconds1761 seconds and 1601 seconds The response time containsthe factors of server continuous rendering response servercontinuous frame-encoding and so forth (more than 70server CPU utilization during interaction) but the networkfactors could be tentatively neglected under the ideal single-user WIFI environment Assuming that we can improvethe serverrsquos processing ability to completely eliminate theserver response delay (perfect 30 fps interaction effect) thetheoretical value of the bandwidth could be 2980 KBps6036 KBps and 3713 KBps respectively In summary theFPS value one of the most important parameters of the 3Dinteractive experience of the studies of [17ndash19]was correlatedreciprocally with the number of simultaneous tasks on theserver side (server task-pipelines) server processing abilityand network bandwidth and an obvious delay of interactionwas clearly presented By contrast all of the Internet deviceschosen to test our method could provide an FPS value at30ndash60 and the FPS valuewas not correlatedwith server tasksserver processing ability and network bandwidth but with
the device GPU (Graphics Processing Unit) Undoubtedlyalmost all GPUs of Internet devices in the market could meetthe demands of our light Slave model rendering Moreoverthe peak value of the bandwidth under the behavior requestfor final image was less than 50KBps and less than 1 KBpsat idle (including the operation of Slave model) thus bothwere far below those tests of methods of [17ndash19] Thusour method could be more suitable for a complex mobileInternet environment Moreover the complex compressiontransmission methods of [17ndash19] were built on customizedsoftware of plug-ins for the client which is hard to applyto pure web pages and falls short of wide compatibility Bycontrast that is our methodrsquos advantage
For the server load because the Master volume in ourmethod only responds to the final behavior request (instruc-tions) instead of the continuous request of the studies of[9 17ndash19] the test showed that server CPUutilizationwas lessthan 20 at idle (including the operation of Slavemodel) andless than 30 for responding to behavior request thus it wasfar below of those tests ofmethods of [17ndash19] with nearly a fullload during all the operation time under the same conditionsof SMI data quality of final required image server config-uration and volume rendering algorithm Theoretically ourplatform can respond to more independent tasks (pipelines)at the same time Table 2 shows that while the server wasrunning four independent tasks simultaneously the responsetime of our method remained at approximately 1 s in the 100-KBps client bandwidth much better than [17ndash19] which waswith more than 16 secondsrsquo delay of end frame in the singletask status In fact because user viewing and processing ofthe 3D medical volume occurred via interaction on the Slavemodel the delay experience wasmuch fewer than that in [17ndash19] as detailed in Section 41That indicates ourmethod couldhave some more advantages including much less networkload and server load to serve much more independent tasksand users
In addition [17ndash19] did not propose an effective inter-active method to clip the 3D volume what our method hasaccomplished
6 Conclusion
This paper proposed aMaster-Slave dual-channel interactionmode which absorbed the advantages of the surface ren-dering of WebGL and remote volume rendering to builda visualization interaction platform to rapidly view andprocess 3Dmedical volume reconstruction in a pure webpageenvironment This platform could be used in a diversifiednetwork environment (100-KBps bandwidth for extreme)and diversified client devices (no special hardware and oper-ating system constraints) without any special installation butwith a good user experience similar to native applicationThese features could serve authorized users to convenientlyaccess SMIs and present 3D visualization anywhere andcould be a technological basis of online communicationamong doctors and patients [34] Or at least it could helpclinicians to arbitrarily acquire 3D image according to theirclinical needs instead of the static pictures that radiologistsdrafted For example clinicians could conveniently observe
12 International Journal of Telemedicine and Applications
the complicated relationship of skull base bone and bloodvessel via this platform (Figure 6(d)) in the past this pieceof information was provided by a few static images generatedfrom radiologistrsquos understanding
Additionally this solution is designed to be followed asanHTML5 standard interface indicating any authorized usercould rapidly access the application via one URL hyperlinkand the platform could be easily integrated into othertelemedical systems seamlessly as a third-party applicationfor example the Supplementary Materials (2 Demonstrationof the operation procedure on laptop linked by emailmp4and 3 Demonstration of the operation procedure on iPadvia websitemp4) These features may play an assistant roleto improve work patterns and extend the range of applicationof traditional medical image systems For example it can beapplied to regional health telemedicine andmedical imagingeducation at a low cost because there is no special cost ofdeploy and upgrade of software network and hardware onthe client side And more importantly a teacher can vividlyshow the students the image characteristics of the diseaseanywhere a clinician can intuitively introduce disease to thepatients anywhere and ask expertsrsquo help anywhere That maypromote the development of mobile medical technology to acertain degree
Conflicts of Interest
There are no conflicts of interest declared
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China (no 81171866) the National Key BasicResearch Program of China (no 2014CB541602) and theSocial Livelihood Science andTechnology Innovation SpecialProject of CSTC (no cstc2015shmszx120002)
References
[1] Q Zhang M Alexander and L Ryner ldquoSynchronized 2D3Doptical mapping for interactive exploration and real-time visu-alization of multi-function neurological imagesrdquo ComputerizedMedical Imaging and Graphics vol 37 no 7-8 pp 552ndash567 2013
[2] S Zimeras and L G Gortzis ldquoInteractive tele-radiological seg-mentation systems for treatment and diagnosisrdquo InternationalJournal of Telemedicine and Applications vol 2012 Article ID713739 p 15 2012
[3] S E Mahmoudi A Akhondi-Asl R Rahmani et al ldquoWeb-based interactive 2D3D medical image processing andvisualization softwarerdquo Computer Methods and Programs inBiomedicine vol 98 no 2 pp 172ndash182 2010
[4] L Faggioni E Neri I Bargellini et al ldquoIPad-based primary2D reading of CT angiography examinations of patients withsuspected acute gastrointestinal bleeding preliminary experi-ence 1Lrdquo British Journal of Radiology vol 88 no 1047 ArticleID 20140477 2015
[5] J B Park H J Choi J H Lee and B S Kang ldquoAn assessment ofthe iPad 2 as a CT teleradiology tool using brain CT with sub-tle intracranial hemorrhage under conventional illuminationrdquoJournal of Digital Imaging vol 26 no 4 pp 683ndash690 2013
[6] P T Johnson S L Zimmerman D Heath et al ldquoThe iPad asa mobile device for CT display and interpretation diagnosticaccuracy for identification of pulmonary embolismrdquo EmergencyRadiology vol 19 no 4 pp 323ndash327 2012
[7] J M Noguera and J R Jimenez ldquoVisualization of very large 3Dvolumes on mobile devices and WebGLrdquo in Proceedings of the20th International Conference in Central Europe on ComputerGraphics Visualization and Computer Vision WSCG 2012 pp105ndash112 June 2012
[8] J Congote A Segura L Kabongo et al ldquoInteractive visu-alization of volumetric data with WebGL in real-timerdquo inProceedings of the 16th International Conference on 3D WebTechnology (Web3D rsquo11) pp 137ndash145 June 2011
[9] A Schiewe M Anstoots and J Kruger ldquoState of the art inmobile volume rendering on iOS devicesrdquo in Proceedings of theEurographics Conference on Visualization (EuroVis rsquo15) 2015
[10] A Evans M Romeo A Bahrehmand et al ldquo3D graphics on theweb a surveyrdquo Computers amp Graphics vol 41 no 1 pp 43ndash612014
[11] M Callieri R M Andrei M D Benedetto et al ldquoVisualizationmethods formolecular studies on theweb platformrdquo inProceed-ings of the 15th international conference on web 3D technology(Web3D rsquo10) pp 117ndash126 July 2010
[12] J Jimenez A M Lopez J Cruz et al ldquoA Web platform forthe interactive visualization and analysis of the 3D fractaldimension of MRI datardquo Journal of Biomedical Informatics vol51 pp 176ndash190 2014
[13] WebGL httpswwwkhronosorgwebgl[14] C R Butson G Tamm S Jain T Fogal and J Kruger
ldquoEvaluation of interactive visualization on mobile computingplatforms for selection of deep brain stimulation parametersrdquoIEEE Transactions on Visualization and Computer Graphics vol19 no 1 pp 108ndash117 2013
[15] M Levoy ldquoDisplay of surfaces from volume datardquo IEEE Com-puter Graphics and Applications vol 8 no 3 pp 29ndash37 1988
[16] E Kotter T Baumann D Jager and M Langer ldquoTechnologiesfor image distribution in hospitalsrdquo European Radiology vol 16no 6 pp 1270ndash1279 2006
[17] S Park W Kim and I Ihm ldquoMobile collaborative medical dis-play systemrdquo Computer Methods and Programs in Biomedicinevol 89 no 3 pp 248ndash260 2008
[18] J R Mitchell P Sharma J Modi et al ldquoA smartphone client-server teleradiology system for primary diagnosis of acutestrokerdquo Journal of Medical Internet Research vol 13 no 2 p e312011
[19] T Hachaj ldquoReal time exploration and management of largemedical volumetric datasets on small mobile devicesmdasheval-uation of remote volume rendering approachrdquo InternationalJournal of Information Management vol 34 no 3 pp 336ndash3432014
[20] P Quax J Liesenborgs A Barzan et al ldquoRemote renderingsolutions using web technologiesrdquoMultimedia Tools and Appli-cations vol 75 no 8 pp 4383ndash4410 2016
[21] D Preuveneers Y Berbers and W Joosen ldquoThe future ofmobile E-health application development exploring HTML5for context-aware diabetes monitoringrdquo Procedia ComputerScience vol 21 pp 351ndash359 2013
[22] J M Noguera J J Jimenez and M C Osuna-Perez ldquoDevel-opment and evaluation of a 3D mobile application for learningmanual therapy in the physiotherapy laboratoryrdquo Computersand Education vol 69 pp 96ndash108 2013
International Journal of Telemedicine and Applications 13
[23] Y S Kang ldquoVolume rendering overviewrdquo inGPU Programmingand Cg Language Primer chapter 14 pp 154-155 1st edition2009
[24] Visualization Toolkit (VTK) httpwwwvtkorg[25] vtkPiecewiseFunction Class Reference of VTK httpwwwvtk
orgdocnightlyhtmlclassvtkPiecewiseFunctionhtml[26] vtkColorTransferFunction Class Reference of VTK http
wwwvtkorgdocnightlyhtmlannotatedhtml[27] vtkPlane Class Reference of VTK httpwwwvtkorgdoc
nightlyhtmlclassvtkPlanehtml[28] L Ma D-X Zhao and Z-Z Yang ldquoA software tool for
visualization of molecular face (VMF) by improving marchingcubes algorithmrdquo Computational and Theoretical Chemistryvol 1028 pp 34ndash45 2014
[29] G F Wang J Liu and W Liu ldquoDIB display and window trans-formation technology for DICOM medical imagesrdquo ChineseJournal of Medical Device vol 18 no 8 pp 1ndash5 2005
[30] L Qiao X Chen L X Yang et al ldquoDevelopment of medi-cal image three-dimensional-visualization platform based onhome network and pure web conditionsrdquo Beijing BiomedicalEngineering vol 34 no 3 pp 229-233286 2015
[31] M Ying and J Miller ldquoRefactoring legacy AJAX applications toimprove the efficiency of the data exchange componentrdquo Journalof Systems and Software vol 86 no 1 pp 72ndash88 2013
[32] HTML5 httpwwww3orgTR2012CR-html5-20121217[33] QIHU 360 SOFTWARE CO LIMITED 360 Internet Security
httpwww360safecomInternet-securityhtml[34] L Qiao Y Li X Chen et al ldquoMedical high-resolution image
sharing and electronic whiteboard system a pure-web-basedsystem for accessing and discussing lossless original images intelemedicinerdquoComputerMethods and Programs in Biomedicinevol 121 no 2 pp 77ndash91 2015
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
4 International Journal of Telemedicine and Applications
(a) Voxels in volume data (b) Pseudo color amp opacity for each voxel (c) Clipping
Figure 2 Schematic diagram of the direct volume rendering method
Size YSize X
Size ZSize Z
Size XSize Y
Y Y
X X
Z Z
(a) Original SMIs
Step 1 Resampling the original SMIs for rst compression
Shrink each slice
(b) Resampling amp strech
(c1) Contour value minus800
(c2) Contour value 200Size 526MBStep 2 Building
3D contour model
(d1) Polygons 50 Polygons 10
(d2) Polygons 50 Polygons 10
Step 3 Reduce the triangles
Step 4Saved asvtk le
(eg 512 times 512 times 495 249 MB) 311MBSize 104MBSize
333MBSize 913MBSize674MBSize
2 times 32 times 495)(eg 3
Figure 3 Schematic diagram of the generation and compression of the Slavemodel
SMIs respectively to acquire a skin model or skeleton model(Figures 3(c1) and 3(c2)) In this case the data size of the skinmodel and skeleton model could be reduced to 674MB and526MB respectively
Step 3 (reducing the total number of polygons for secondslimming) The 3D geometrical contour model from Step 2is constitutive of polygons and we could reduce the totalnumber of polygons to further reduce the data size In thiscase the number was reduced to 50 and 10 and thedata sizes were reduced to approximately 3MB and 1MBrespectively (Figures 3(d1) and 3(d2))
Step 4 (saved as vtk model file) The slimmed model wouldbe saved in a file of the vtk format which is a storage formatof the geometrical model supported by WebGL and thenstored in the same directory with original SMIs
The result of Figures 3(d1) and 3(d2) preliminarily showedthat the Slave model whose data size is approximately 1MBstill could be fit for interactive navigation (more experimentsdetailed in Section 41) Moreover an additional mutualinterface of the client was designed as a remote parametersetting of the contour value and the total number of polygonsto permit users to customize the Slavemodel as desired
232 Design and Quantization of the Interactive Behaviors inthe Slave Model For a friendly interactive experience on theclient side a simple style of interactive behaviors of the clientand corresponding quantitative methods should be designed
to guide remote volume rendering including the viewingangle request and clipping request
For the Viewing Angle Request The interaction of viewingangle includes rotation flipping zooming and panning [30]We defined the interactive 3D scene including only one Slavemodel and one camera and customized an interactive modeof the camera in VR technology to meet our requirementFigure 4(a) shows the principle of the interaction of theviewing angle between the camera and Slave model Welimited the lens of the camera to always be taken on point1198751 (0 0 0) which was the origin of the world coordinatesystem The interactive behaviors of the viewing angle werecarried out by only changing the camera position exceptpanning For panning it was carried out by changing theSlavemodel position see examples in Figure 4(b)
According to the definition while sending requestinstructions to the server for remote volume rendering theinstructions only include three quantization parameters fromthe client to accurately describe the interactive behavior ofthe viewing angle request to the Slavemodel The three quan-tization parameters include the following attribute view updirection attribute position of camera and attribute positionof the center point of the Slavemodel
Here attribute view up direction of camera indicates thedirection of the top of the camera For example (0 minus1 0)a value of the View Up direction indicates the camera isperpendicular to the XZ plane in the world coordinatesystem and the top of the camera is in the opposite direction
International Journal of Telemedicine and Applications 5
(a) Principle of interaction between camera and Slavemodel
(i) Original camera view
(iii) Zoom via shorteningthe distance from CaA to P1
(iv) Flip from CaA to CaB (v) Panning via movingthe center point of Slavemodel from P1 to P2
(ii) Rotate around axis CaA-P1 (Y-axis)
(b) Example of interactions of viewing angle request
Figure 4 Principle of the interaction of the viewing angle between the camera and Slavemodel and the corresponding example
along the 119885-axis (see the relationship between camera CaAand Slave model in Figure 4(a) and corresponding example(i) original camera view in Figure 4(b))
In fact WebGL has provided some interaction methodsof the moving camera to view 3D objects via mouse- andgesture-based interactive behaviors which were packagedinto a library of TrackballControlsjs for a web page scriptWe could directly utilize the js library to acquire the threequantization parameters However we need to rewrite thejs library to restrict and customize the interactive modeof the camera that we have defined to ensure the consis-tency of the interactive rule between the Master and Slaveside
For the Clipping Request For arbitrarily clipping and viewingthe internal anatomy structure of theMaster volume a greenrectangle object named the Clipping plane was added in theabove defined interactive 3D scene to allow users to mimicthe clipping behavior in the Slavemodel see Figures 5(a) and5(b) Here a virtual control panel was designed separately tosupport the interactive behaviors of users using the Clippingplane (Figure 5(c))
During the initialization phase the Clipping plane coversthe 119883119884 plane in the world coordinate system (Figure 5(a))Its specific location could be accurately described by thefollowing two quantization parameters in the computergraphics method normal vector from the origin point tothe Clipping plane and the coordinate position of a pointon the Clipping plane Here its default normal vector was0 0 1 one point 119875 on the plane was on the position(0 0 0)
For intuitive operation of the user experience the interac-tive behaviors regarding the Clipping plane include rotationand panning and both can be operated by a virtual controlpanel However for the accurate description of the requestinstructions for remote volume rendering the two quantiza-tion parameters should be used
Therefore formula (1) was deduced to calculate thenormal vector from the angles of rotation
119899119900119903119898119886119897 V119890119888119905119900119903 = sin (120579119910) minus sin (120579119909)
sdot cos (120579119910) cos (120579119909) sdot cos (120579119910) (1)
Here 120579119909 120579119910 are the final rotation angles that rotate theClipping plane clockwise around the 119883-axis and 119884-axisrespectively
For example in Figure 5(b) a user rotated the Clippingplane 30 degrees clockwise around the 119884-axis and minus15degrees clockwise around the 119883-axis and then panned theClipping plane 95 pixels in the positive direction along the119885-axis While submitting for remote volume rendering 120579119909 =minus15 120579119910 = 30 the normal vector was 05 02241 08365according to formula (1) Meanwhile according to the char-acteristic of monolithic translation the one point 119875 on theClipping plane was moved to position (0 0 95) That wouldbe the two quantization parameters of request instructions forclipping
24 Remote Interaction between the Server and Client Theinteractive behaviors of the Slavemodel include viewing anglerequest and clipping request Moreover coloropacity request isoperated by the dropdown list of the preset schemes (detailedin Section 22) All of these requests would be submitted tothe server in the form of quantization parameters which arenamed as request instructions
However in addition to responding to the clientrsquos instruc-tions the server has to find an efficient way to manageand allocate system resources to enhance the robustness ofthe high frequency of interaction among multiple users forexample to reduce the probability of deadlock cumulativedelay and task-pipeline crossing
In our system we used relational database schema tomanage the request instructions from the client and to
6 International Journal of Telemedicine and Applications
Interactiveeect
Clippingplane
XY
Z
PX
Y
Z
(a) Initial state of Clipping plane
PX
Y
Z
(i) Rotated 30 degrees (ii) Further rotated minus15 degreesaround the X-axis
(ii) Further panning 95
P
XY
Z
P
XY
Z
Y-axisaround the Z-axisalong the
(b) Instances of interaction of Clipping plane (c) Control panelof Clipping plane
Figure 5 Design of the interaction of clipping to the Slavemodel and corresponding example
allocate the system resources for guiding 3D rendering Thecore tables and their data storage structure are designed inTable 1
According to Table 1 if a user chooses a group of SMIsfrom a list derived from the base table the server wouldquery the pipeline table whether the SMIs is serving theclient If No the SMIs would be loaded in memory anda task-pipeline between the user and SMIs would be builtThereafter the state of the pipeline table would be changed toldquoYesrdquoWhile being Yes the client would be permitted tomakethe next interactive request After that for each request fromthe client the servermust query the state of the behavior tablewhether the pipeline is ldquopendingrdquo or ldquoprocessingrdquo If No theserver would accept the request with a record of a ldquopendingrdquostate Additionally the listener on the server would real-timearrange these pending requests to the server rendering inqueue While a request is being rendered the correspondingstate of behavior table would be changed as ldquoprocessingrdquo toprevent malignant submission While the job is finished thestate would be changed to ldquofinishedrdquo and a 2D projectionimage would be generated and saved in the specified HTTPsharing path
The client adopted AJAX (Asynchronous JavaScript andXML) [31] to dynamically query the state of behavior tablefrom the server at a fixed frequency of 300ms While render-ing was finished the client would download and update the2D projection image via the HTML + js web page technique
3 System Description
31 System Architecture While Figure 1 depicts the Master-Slave dual-channel interaction mode it also presents theoverall architecture of the final platform Here we adoptedthe browserserver architecture for which the communi-cation with each other was via the HTTP protocol Theserver side consists of the Listener amp Request ManagementServer (in zone B) and Rendering Server (in zone C) Theformer accepts and manages the requests from the client
and guides the Rendering Server to an orderly responseThe latter renders the SMIs and generates a 2D projec-tion image for the client downloading via HTTP At thesame time the client side was designed by the standardof HTML5 [32] We used ltimggt an HTML image tagto display the 2D projection image in the form of pseudo3D and used AJAX + js technology to realize flicker-freepage updates and client RIA (rich Internet application)over diversified web browsers without installation More-over HTML5 includes the WebGL standard and the Slavemodel could run on any HTML5-supporting web browserswithout any web plug-ins or special software tools Thusthe whole platform system could run under mature networktechnologies
32 System Front-End System front-end is anHTML5-basedpure website that provides entrance to SMI access and 3Dpresentation The users require a username and password toenter the system or any experts could be invited to enter aspecific task-pipeline with one URL hyperlink from a helpseeker In brief using an HTML5-supporting web browserincluding the major versions of Firefox chrome Operasafari IE 10+ the 3D volume can be viewed and processedas in Figure 6 and the Supplementary Materials availableonline at httpsdoiorg10115520174074137 (1 Demon-stration of the operation procedure on laptop via websitemp4)
Here the interactive behaviors of the user include theviewing angle clipping and coloropacity transformationThefirst two items were operated on the Slave model which wasreconstructed on the client side While the user wants toobserve the final required image of high quality the userjust presses the submit button to submit the current behaviorrequest instruction to the server and waits for the final 2Dprojection image to be updated
Figure 6(a) presents the initial state of the Slave model(right) and 2D projection image (left) In Figure 6(b) mouseor hand gesture is used to operate the Slave model to
International Journal of Telemedicine and Applications 7
Table1Re
lationald
atas
torage
structureo
frequestbehaviorsfrom
clientand
resource
allocatio
non
thes
erver
Table
Field
Note
Basetable
DIC
OMs_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
grou
pof
SMIs
Torecord
theb
asicinform
ationof
SMIs
onthes
erverside
Storagep
ath
Thes
torage
path
ofSM
Ison
thes
erver
side
Pipelin
etable
pipelin
e_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
pipelin
efor
view
ingag
roup
ofSM
Isbetweenthec
lient
andserver
Basedon
basetableto
record
the
conn
ectedrelationshipbetweentheS
MIs
rend
eringin
mem
oryon
thes
ervera
ndremoteu
sero
nthec
lient
DIC
OMs_ID
FORE
IGNKE
Yof
basetableto
decla
rewhich
grou
pof
SMIsthec
lient
isview
ing
Client
Client
user
IPaddress
IPaddresso
fthe
committer
State
Whether
theS
MIsfin
ished
rend
eringin
mem
oryon
thes
erverYesallowthe
clienttonext
operation
Noprompt
for
waitin
g
Behavior
table
behavior_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofeach
behavior
requ
estevent
from
the
client
Basedon
pipelin
etable
torecord
operationrequ
estsfro
mthec
lient
user
andtheirp
rocesssta
te
pipelin
e_ID
FORE
IGNKE
Yof
pipelin
etable
todecla
rewhich
pipelin
eonview
ingand
processin
g
Behavior
type
Inclu
ding
view
inga
ngle
clippingor
coloropacity
transfo
rmation
Behavior-in
structio
n
From
thec
lienttotransfe
rthe
parametersinterm
softhe
correspo
nding
behavior
typerecordedin
theform
ofqu
antizationparameters
State
Pend
ingprocessin
gor
finish
ed
8 International Journal of Telemedicine and Applications
(a) Initial state (b) Viewing angle interaction
(c) Clipping interaction (d) Zooming and coloropacity interaction
Figure 6 Screenshot and operational demonstration
transform the viewing angle from Figure 6(a) to Figure 6(b)(red solid box) and then the submit button is pressed to sendthe behavior request instruction to the server to require anupdated projection image from the Master volume (yellowsolid box) In Figure 6(c) using virtual control panel (reddashed box) to operate Clipping plane after submitting thenew projection image was then updated (left) In Figure 6(d)the mouse wheel or hand gesture is used to zoom the Slavemodel and then the coloropacity theme is submitted via thedropdown list (bottom right corner) to guide remote volumerendering
4 Mode of Availability of the System
In this paper we designed aMaster-Slave dual-channel inter-action mode to improve a problem of remote visualizationinteraction of the volume reconstruction of SMIs whichusually have a huge data size and calculation cost Finallya visualization interaction platform has been built Particu-larly there is no special software and hardware requirementfor the clients and no special network environment isneeded
Tomeasure whether the effectiveness and performance toadapt to the mobile Internet application has characteristicsof lightweight three radiologists with more than 3 years ofexperience were invited to test the interactive experienceimaging quality and compatibility over diversified Internetdevices Moreover four junior students majoring in Biomed-ical Engineeringwere invited to test and quantize the networkload and response time
TestingData 3 groups of SMIs from the radiology departmentof Southwest Hospital of Chongqing include the following
(1) head and neck CT with 495 slices 512 lowast 512 2494MB(2) trunk CT with 609 slices 512 lowast 512 3065MB and (3)thorax andmandible CT with 500 slices 512lowast512 2516MBServer An IBM X3650M4 workstation (CPU 4 times E5-2603RAM 16GB OS Windows Server 2008) was connectedto the Internet via 10Mbps The 2D projection image wasset as the resolution of 512 lowast 512 with high quality JPEGcompression (level-8)
Client 4 Internet devices were located at the 3G or WIFImobile Internet with the server In Figure 7(a) there is aFounder E520 personal computer running on winXP withFirefox 46 web browser via 3G usb card In Figure 7(b) thereis an Acer E15 laptop running on win7 with chrome 43 webbrowser at WIFI In Figure 7(c) there is a HuaWei Honor6 smart phone running on EMUI 30 which is developedfrom Android 442 with the Opera 37 web browser at 3Gnet In Figure 7(d) there is an iPad mini running on IOS6with safari at WIFI All of them could be representative inthe mobile Internet Additionally each was limited to a 100-KBps wireless bandwidth by 360 security firewall for extremetest [33]
41 Interactive Experience Imaging Quality and Compatibil-ity To ensure the consistency of the context the testing data(1) were considered as the main object of reference Here thedemonstration and effect of the testing data (1) are shown inFigure 6
Figure 6 shows that this Slave model of the testing data(1) which was compressed to 104MB could be fit to thedemands of interactive navigation to accurately guide remotevolume rendering and shows that the projection image couldbe displayed to keep the original quality
International Journal of Telemedicine and Applications 9
(a) Personal computer (b) Laptop computer
(c) Smartphone (d) iPad
Figure 7 The system is accessed via four different devices and web browsers under extreme bandwidth
For further validation three radiologists were invited totest the platform with the similar behaviors in Figure 6 Thisjob was repeated to operate three groups of different testingdata through different web browsers and Internet devicesthat were all aforementioned Additionally a five-pointscale was designed to respond to the radiologistsrsquo requests(Table 2)
Table 2 presents a very positive evaluation from the statis-tical result of the imaging quality and interactive experienceand all three radiologists expressed that is an interestingway whether as a gesture-based or mouse-based interactionNo installations were required from the client and goodcompatibility was obtained (same result is shown in Table 3)However one radiologist thought the default Slave modelof the testing data (1) with a skeleton model of 104MB(Figure 3(d2)) was a little rough He suggested to use theskeleton model of 526MB or skin model of 913 KB (Figures3(c1) and 3(c2)) instead of the 104 model both of which arethe alternatives from the mutual interface for the client asdetailed in Section 231 Additionally all three radiologistsclaimed that the response time for final high quality image isalways about 1 second that is consistent with the test resultin Section 42 However they did accept the performancebecause they could directly operate the Slavemodel smoothlywithout any delay on client side thus a short wait for finalhigh quality image had little influence Moreover one of theradiologists specifically expressed the native application of3D rendering on desktop PC also needs 1 or 2 secondsrsquo delayfor final image this performance of the solution running onsmart phone is similar to the native application on PC and isvery good
42 Response Time and Network Load between the Clientand Server The response time can be defined as the timeinterval between the user sending a behavior request tothe system and web browser loading the response resultscompletely The network load can be defined as the amountof data transmission of each interactive behavior Four juniorstudents majoring in Biomedical Engineering were invitedto simultaneously test the platform for one hour During thetest all of the students respectively viewed and processed thetesting data (1) through the platform and performed similarbehaviors to those detailed in Figure 6This test was repeatedby each student with different web browsers and Internetdevices all of which were aforementioned Here we trainedthe students to record the real-time network traffic by the 360security firewall and the time interval by the script log Theresults are showed in Table 3
In Table 3 for each device the amount of data trans-mission and response time of the first loading are muchhigher than the behavior request and idle status because thefirst loading asks the client to download the necessary Slavemodel The response time of the first loading is lt12 s at 100-KBps client bandwidth which is similar to the total timespent on Slavemodel downloading In fact the response timeof first loading also includes the cost time of SMI loadingand rendering on the server side but the server monitorpresents the cost time as lt7 s because SMI loading and Slavemodel downloading are simultaneous and this cost could beneglected
After the first loading the network load of the behaviorrequest for final image is spent on request instructions sub-mission final projection image downloading and the serverprocessing status querying at a fixed frequency of 300ms
10 International Journal of Telemedicine and Applications
Table 2 Statistic results of the five-point scale concerning the interactive experience and imaging quality
Focus Subject Votes from five-point scaleVeryagree Agree Uncertainty Disagree Very
disagree
Interactive experienceMouse-basedGesture-based 998771
The operation of viewing and processing isaccurate
3998771 3
Simple and easy to use without training 2 1998771 3
A good user experience similar to nativeapplication
2 1998771 2 1
Imaging quality
Slavemodel could meet demands of navigation 2 12D image fromMaster volume could provide highquality reconstruction result via utilizing thedisplay capability of Internet terminals
3
Others Latency time for final high quality images 3
Table 3 Average response time and amount of data transmission from the different behaviors and devices at 100-KBps
Devices
First loading (averagevalue)
Behavior request for finalimage (average value)
Idle status ofbandwidthoccupation(KBps)
Frames PerSecond (FPS)Amount of
data trans (KB)Response time(seconds)
Amount ofdata trans (KB)
Response time(seconds)
PC (3G net) 108544 1123 4710 112 023 48ndash60Laptop (WIFI) 106496 1051 4608 089 031 48ndash60iPad mini (WIFI) 107941 1093 4403 095 035 50ndash60Smartphone (3G net) 107502 1072 3994 102 028 30ndash40
where the projection image downloading is the main partThe results present a complete amount of data transmission ofeach behavior request is close to 50KB but the correspondingresponse time for final image is approximately 1 s in 100KBpsThat is because besides the network load the rest of timeis spent on server processing and time interval of querying(300ms) The short delay could be acceptable for practicalapplications as detailed in Section 41
While in the idle status including the userrsquos interactivebehavior (operation) on the Slave model the bandwidthoccupation is almost negligible Moreover there are no sig-nificant differences among diversified devices and networksat 100-KBps client bandwidth
Additionally the FPS of Slave model rendering on theclient side is measured via statsminjs which is a librarypackaged from WebGL When the client rendering is com-pleted FPS could remain at the upper limit
5 Discussion and Comparison
Due to the nonprivate network environment and diversifiedpurposes application software on one device must competefor limited resources of the network and computing Thuslightweight which requires lower resources of the networkand computing and avoids unnecessary installation could bean important characteristic of mobile Internet applicationAdditionally to avoid compatibility problems caused by dif-ferent operating systems and client hardware characteristics
using HTML5 and related web technologies running onmajor web browsers may be a choice to adapt to cross-platform deployment and the characteristics of lightweight[21]
Therefore this article described research on a methodfor the Master-Slave dual-channel interaction mode whichis based on the HTML5 standard (also a wider web com-patibility standard of WebGL) to permit users viewing andprocessing million-megabyte-class SMIs via different clientdevices (without hardware or operating system constraints)over a pure web page (without installing special software) andin a diversified network environment
Compared with the traditional PACS mode of ldquoimagecompression transmission local reconstructionrdquo becauseour method does not transfer original SMIs to the clientit neither needs to wait for a long time to download norrequests special demands of the software environment onthe client device Moreover compared with the state ofthe art of local volume rendering based on WebGL andHTML5 technology [7ndash9] our methods have no specialdemands of client hardware for computation that may reducebattery consumption [9] and enhance the compatibility ofthe client devices our methods may be more fit for mobileInternet application Additionally the client does not accessthe original SMIs which could be much safer
Compared with popular solutions based on absolutesurface rendering technology [11 12 14] our method haslocally used this technique to only support light Slavemodel
International Journal of Telemedicine and Applications 11
for interactive navigation therefore the demands of thenetwork bandwidth and device rendering capability arealmost negligible in ourmethodMoreover the final requiredimage of high quality is a projection drawing from volumerendering technology indicating that our method could beconvenient for users to view the internal anatomy structureand complex organization relationship however surfacerendering technology could not support these characteristicsof volume
Compared with solutions based on absolute remote vol-ume rendering methods our method adopted the Master-Slave dual-channel interaction mode to strengthen the userexperience which is an entirely different approach from themode of ldquocontinuous request continuous responserdquo of studiesof [9] Although studies of [17ndash19] proposed several methodsto improve the mode to reduce the network load includingvideo-compressed transmission [17] tile-based transmission[18] and variable resolution transmission [19] the pressureof server load and real-time data stream increased We usedthe same SMI data (testing data (1)) the same quality offinal required image and the same server configurationand volume rendering algorithm in Section 4 to test themethods in [17ndash19] For quantitative test we formulated aninteractive trajectory with fixed 30 request points theoreti-cally assuming 30 rendering requests triggered per 1 secondthere should be 30 fps render effects On this basis [17]adopted 512 lowast 512 mpeg-4 compression for transmission[18] adopted 8 lowast 8 blocks to segment each 512 lowast 512 frameand transfer the blocks which are different from previousframe in the process of interaction and [19] adopted 64 lowast64 resolution transmission in the process of interaction but512 lowast 512 for end frame The test was repeated for 5 timesunder the condition of one single client monopolizing thebandwidth and server under ideal WIFI environment Theresults showed that the total amount of data transmissionfor each interaction test of [17ndash19] was 2980 KB 6036 KBand 3713 KB respectively the average response time of firstframe was 110 seconds 059 seconds and 052 seconds andthe average cumulative response time of end frame (30thframe which is the final required image) was 2131 seconds1761 seconds and 1601 seconds The response time containsthe factors of server continuous rendering response servercontinuous frame-encoding and so forth (more than 70server CPU utilization during interaction) but the networkfactors could be tentatively neglected under the ideal single-user WIFI environment Assuming that we can improvethe serverrsquos processing ability to completely eliminate theserver response delay (perfect 30 fps interaction effect) thetheoretical value of the bandwidth could be 2980 KBps6036 KBps and 3713 KBps respectively In summary theFPS value one of the most important parameters of the 3Dinteractive experience of the studies of [17ndash19]was correlatedreciprocally with the number of simultaneous tasks on theserver side (server task-pipelines) server processing abilityand network bandwidth and an obvious delay of interactionwas clearly presented By contrast all of the Internet deviceschosen to test our method could provide an FPS value at30ndash60 and the FPS valuewas not correlatedwith server tasksserver processing ability and network bandwidth but with
the device GPU (Graphics Processing Unit) Undoubtedlyalmost all GPUs of Internet devices in the market could meetthe demands of our light Slave model rendering Moreoverthe peak value of the bandwidth under the behavior requestfor final image was less than 50KBps and less than 1 KBpsat idle (including the operation of Slave model) thus bothwere far below those tests of methods of [17ndash19] Thusour method could be more suitable for a complex mobileInternet environment Moreover the complex compressiontransmission methods of [17ndash19] were built on customizedsoftware of plug-ins for the client which is hard to applyto pure web pages and falls short of wide compatibility Bycontrast that is our methodrsquos advantage
For the server load because the Master volume in ourmethod only responds to the final behavior request (instruc-tions) instead of the continuous request of the studies of[9 17ndash19] the test showed that server CPUutilizationwas lessthan 20 at idle (including the operation of Slavemodel) andless than 30 for responding to behavior request thus it wasfar below of those tests ofmethods of [17ndash19] with nearly a fullload during all the operation time under the same conditionsof SMI data quality of final required image server config-uration and volume rendering algorithm Theoretically ourplatform can respond to more independent tasks (pipelines)at the same time Table 2 shows that while the server wasrunning four independent tasks simultaneously the responsetime of our method remained at approximately 1 s in the 100-KBps client bandwidth much better than [17ndash19] which waswith more than 16 secondsrsquo delay of end frame in the singletask status In fact because user viewing and processing ofthe 3D medical volume occurred via interaction on the Slavemodel the delay experience wasmuch fewer than that in [17ndash19] as detailed in Section 41That indicates ourmethod couldhave some more advantages including much less networkload and server load to serve much more independent tasksand users
In addition [17ndash19] did not propose an effective inter-active method to clip the 3D volume what our method hasaccomplished
6 Conclusion
This paper proposed aMaster-Slave dual-channel interactionmode which absorbed the advantages of the surface ren-dering of WebGL and remote volume rendering to builda visualization interaction platform to rapidly view andprocess 3Dmedical volume reconstruction in a pure webpageenvironment This platform could be used in a diversifiednetwork environment (100-KBps bandwidth for extreme)and diversified client devices (no special hardware and oper-ating system constraints) without any special installation butwith a good user experience similar to native applicationThese features could serve authorized users to convenientlyaccess SMIs and present 3D visualization anywhere andcould be a technological basis of online communicationamong doctors and patients [34] Or at least it could helpclinicians to arbitrarily acquire 3D image according to theirclinical needs instead of the static pictures that radiologistsdrafted For example clinicians could conveniently observe
12 International Journal of Telemedicine and Applications
the complicated relationship of skull base bone and bloodvessel via this platform (Figure 6(d)) in the past this pieceof information was provided by a few static images generatedfrom radiologistrsquos understanding
Additionally this solution is designed to be followed asanHTML5 standard interface indicating any authorized usercould rapidly access the application via one URL hyperlinkand the platform could be easily integrated into othertelemedical systems seamlessly as a third-party applicationfor example the Supplementary Materials (2 Demonstrationof the operation procedure on laptop linked by emailmp4and 3 Demonstration of the operation procedure on iPadvia websitemp4) These features may play an assistant roleto improve work patterns and extend the range of applicationof traditional medical image systems For example it can beapplied to regional health telemedicine andmedical imagingeducation at a low cost because there is no special cost ofdeploy and upgrade of software network and hardware onthe client side And more importantly a teacher can vividlyshow the students the image characteristics of the diseaseanywhere a clinician can intuitively introduce disease to thepatients anywhere and ask expertsrsquo help anywhere That maypromote the development of mobile medical technology to acertain degree
Conflicts of Interest
There are no conflicts of interest declared
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China (no 81171866) the National Key BasicResearch Program of China (no 2014CB541602) and theSocial Livelihood Science andTechnology Innovation SpecialProject of CSTC (no cstc2015shmszx120002)
References
[1] Q Zhang M Alexander and L Ryner ldquoSynchronized 2D3Doptical mapping for interactive exploration and real-time visu-alization of multi-function neurological imagesrdquo ComputerizedMedical Imaging and Graphics vol 37 no 7-8 pp 552ndash567 2013
[2] S Zimeras and L G Gortzis ldquoInteractive tele-radiological seg-mentation systems for treatment and diagnosisrdquo InternationalJournal of Telemedicine and Applications vol 2012 Article ID713739 p 15 2012
[3] S E Mahmoudi A Akhondi-Asl R Rahmani et al ldquoWeb-based interactive 2D3D medical image processing andvisualization softwarerdquo Computer Methods and Programs inBiomedicine vol 98 no 2 pp 172ndash182 2010
[4] L Faggioni E Neri I Bargellini et al ldquoIPad-based primary2D reading of CT angiography examinations of patients withsuspected acute gastrointestinal bleeding preliminary experi-ence 1Lrdquo British Journal of Radiology vol 88 no 1047 ArticleID 20140477 2015
[5] J B Park H J Choi J H Lee and B S Kang ldquoAn assessment ofthe iPad 2 as a CT teleradiology tool using brain CT with sub-tle intracranial hemorrhage under conventional illuminationrdquoJournal of Digital Imaging vol 26 no 4 pp 683ndash690 2013
[6] P T Johnson S L Zimmerman D Heath et al ldquoThe iPad asa mobile device for CT display and interpretation diagnosticaccuracy for identification of pulmonary embolismrdquo EmergencyRadiology vol 19 no 4 pp 323ndash327 2012
[7] J M Noguera and J R Jimenez ldquoVisualization of very large 3Dvolumes on mobile devices and WebGLrdquo in Proceedings of the20th International Conference in Central Europe on ComputerGraphics Visualization and Computer Vision WSCG 2012 pp105ndash112 June 2012
[8] J Congote A Segura L Kabongo et al ldquoInteractive visu-alization of volumetric data with WebGL in real-timerdquo inProceedings of the 16th International Conference on 3D WebTechnology (Web3D rsquo11) pp 137ndash145 June 2011
[9] A Schiewe M Anstoots and J Kruger ldquoState of the art inmobile volume rendering on iOS devicesrdquo in Proceedings of theEurographics Conference on Visualization (EuroVis rsquo15) 2015
[10] A Evans M Romeo A Bahrehmand et al ldquo3D graphics on theweb a surveyrdquo Computers amp Graphics vol 41 no 1 pp 43ndash612014
[11] M Callieri R M Andrei M D Benedetto et al ldquoVisualizationmethods formolecular studies on theweb platformrdquo inProceed-ings of the 15th international conference on web 3D technology(Web3D rsquo10) pp 117ndash126 July 2010
[12] J Jimenez A M Lopez J Cruz et al ldquoA Web platform forthe interactive visualization and analysis of the 3D fractaldimension of MRI datardquo Journal of Biomedical Informatics vol51 pp 176ndash190 2014
[13] WebGL httpswwwkhronosorgwebgl[14] C R Butson G Tamm S Jain T Fogal and J Kruger
ldquoEvaluation of interactive visualization on mobile computingplatforms for selection of deep brain stimulation parametersrdquoIEEE Transactions on Visualization and Computer Graphics vol19 no 1 pp 108ndash117 2013
[15] M Levoy ldquoDisplay of surfaces from volume datardquo IEEE Com-puter Graphics and Applications vol 8 no 3 pp 29ndash37 1988
[16] E Kotter T Baumann D Jager and M Langer ldquoTechnologiesfor image distribution in hospitalsrdquo European Radiology vol 16no 6 pp 1270ndash1279 2006
[17] S Park W Kim and I Ihm ldquoMobile collaborative medical dis-play systemrdquo Computer Methods and Programs in Biomedicinevol 89 no 3 pp 248ndash260 2008
[18] J R Mitchell P Sharma J Modi et al ldquoA smartphone client-server teleradiology system for primary diagnosis of acutestrokerdquo Journal of Medical Internet Research vol 13 no 2 p e312011
[19] T Hachaj ldquoReal time exploration and management of largemedical volumetric datasets on small mobile devicesmdasheval-uation of remote volume rendering approachrdquo InternationalJournal of Information Management vol 34 no 3 pp 336ndash3432014
[20] P Quax J Liesenborgs A Barzan et al ldquoRemote renderingsolutions using web technologiesrdquoMultimedia Tools and Appli-cations vol 75 no 8 pp 4383ndash4410 2016
[21] D Preuveneers Y Berbers and W Joosen ldquoThe future ofmobile E-health application development exploring HTML5for context-aware diabetes monitoringrdquo Procedia ComputerScience vol 21 pp 351ndash359 2013
[22] J M Noguera J J Jimenez and M C Osuna-Perez ldquoDevel-opment and evaluation of a 3D mobile application for learningmanual therapy in the physiotherapy laboratoryrdquo Computersand Education vol 69 pp 96ndash108 2013
International Journal of Telemedicine and Applications 13
[23] Y S Kang ldquoVolume rendering overviewrdquo inGPU Programmingand Cg Language Primer chapter 14 pp 154-155 1st edition2009
[24] Visualization Toolkit (VTK) httpwwwvtkorg[25] vtkPiecewiseFunction Class Reference of VTK httpwwwvtk
orgdocnightlyhtmlclassvtkPiecewiseFunctionhtml[26] vtkColorTransferFunction Class Reference of VTK http
wwwvtkorgdocnightlyhtmlannotatedhtml[27] vtkPlane Class Reference of VTK httpwwwvtkorgdoc
nightlyhtmlclassvtkPlanehtml[28] L Ma D-X Zhao and Z-Z Yang ldquoA software tool for
visualization of molecular face (VMF) by improving marchingcubes algorithmrdquo Computational and Theoretical Chemistryvol 1028 pp 34ndash45 2014
[29] G F Wang J Liu and W Liu ldquoDIB display and window trans-formation technology for DICOM medical imagesrdquo ChineseJournal of Medical Device vol 18 no 8 pp 1ndash5 2005
[30] L Qiao X Chen L X Yang et al ldquoDevelopment of medi-cal image three-dimensional-visualization platform based onhome network and pure web conditionsrdquo Beijing BiomedicalEngineering vol 34 no 3 pp 229-233286 2015
[31] M Ying and J Miller ldquoRefactoring legacy AJAX applications toimprove the efficiency of the data exchange componentrdquo Journalof Systems and Software vol 86 no 1 pp 72ndash88 2013
[32] HTML5 httpwwww3orgTR2012CR-html5-20121217[33] QIHU 360 SOFTWARE CO LIMITED 360 Internet Security
httpwww360safecomInternet-securityhtml[34] L Qiao Y Li X Chen et al ldquoMedical high-resolution image
sharing and electronic whiteboard system a pure-web-basedsystem for accessing and discussing lossless original images intelemedicinerdquoComputerMethods and Programs in Biomedicinevol 121 no 2 pp 77ndash91 2015
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of Telemedicine and Applications 5
(a) Principle of interaction between camera and Slavemodel
(i) Original camera view
(iii) Zoom via shorteningthe distance from CaA to P1
(iv) Flip from CaA to CaB (v) Panning via movingthe center point of Slavemodel from P1 to P2
(ii) Rotate around axis CaA-P1 (Y-axis)
(b) Example of interactions of viewing angle request
Figure 4 Principle of the interaction of the viewing angle between the camera and Slavemodel and the corresponding example
along the 119885-axis (see the relationship between camera CaAand Slave model in Figure 4(a) and corresponding example(i) original camera view in Figure 4(b))
In fact WebGL has provided some interaction methodsof the moving camera to view 3D objects via mouse- andgesture-based interactive behaviors which were packagedinto a library of TrackballControlsjs for a web page scriptWe could directly utilize the js library to acquire the threequantization parameters However we need to rewrite thejs library to restrict and customize the interactive modeof the camera that we have defined to ensure the consis-tency of the interactive rule between the Master and Slaveside
For the Clipping Request For arbitrarily clipping and viewingthe internal anatomy structure of theMaster volume a greenrectangle object named the Clipping plane was added in theabove defined interactive 3D scene to allow users to mimicthe clipping behavior in the Slavemodel see Figures 5(a) and5(b) Here a virtual control panel was designed separately tosupport the interactive behaviors of users using the Clippingplane (Figure 5(c))
During the initialization phase the Clipping plane coversthe 119883119884 plane in the world coordinate system (Figure 5(a))Its specific location could be accurately described by thefollowing two quantization parameters in the computergraphics method normal vector from the origin point tothe Clipping plane and the coordinate position of a pointon the Clipping plane Here its default normal vector was0 0 1 one point 119875 on the plane was on the position(0 0 0)
For intuitive operation of the user experience the interac-tive behaviors regarding the Clipping plane include rotationand panning and both can be operated by a virtual controlpanel However for the accurate description of the requestinstructions for remote volume rendering the two quantiza-tion parameters should be used
Therefore formula (1) was deduced to calculate thenormal vector from the angles of rotation
119899119900119903119898119886119897 V119890119888119905119900119903 = sin (120579119910) minus sin (120579119909)
sdot cos (120579119910) cos (120579119909) sdot cos (120579119910) (1)
Here 120579119909 120579119910 are the final rotation angles that rotate theClipping plane clockwise around the 119883-axis and 119884-axisrespectively
For example in Figure 5(b) a user rotated the Clippingplane 30 degrees clockwise around the 119884-axis and minus15degrees clockwise around the 119883-axis and then panned theClipping plane 95 pixels in the positive direction along the119885-axis While submitting for remote volume rendering 120579119909 =minus15 120579119910 = 30 the normal vector was 05 02241 08365according to formula (1) Meanwhile according to the char-acteristic of monolithic translation the one point 119875 on theClipping plane was moved to position (0 0 95) That wouldbe the two quantization parameters of request instructions forclipping
24 Remote Interaction between the Server and Client Theinteractive behaviors of the Slavemodel include viewing anglerequest and clipping request Moreover coloropacity request isoperated by the dropdown list of the preset schemes (detailedin Section 22) All of these requests would be submitted tothe server in the form of quantization parameters which arenamed as request instructions
However in addition to responding to the clientrsquos instruc-tions the server has to find an efficient way to manageand allocate system resources to enhance the robustness ofthe high frequency of interaction among multiple users forexample to reduce the probability of deadlock cumulativedelay and task-pipeline crossing
In our system we used relational database schema tomanage the request instructions from the client and to
6 International Journal of Telemedicine and Applications
Interactiveeect
Clippingplane
XY
Z
PX
Y
Z
(a) Initial state of Clipping plane
PX
Y
Z
(i) Rotated 30 degrees (ii) Further rotated minus15 degreesaround the X-axis
(ii) Further panning 95
P
XY
Z
P
XY
Z
Y-axisaround the Z-axisalong the
(b) Instances of interaction of Clipping plane (c) Control panelof Clipping plane
Figure 5 Design of the interaction of clipping to the Slavemodel and corresponding example
allocate the system resources for guiding 3D rendering Thecore tables and their data storage structure are designed inTable 1
According to Table 1 if a user chooses a group of SMIsfrom a list derived from the base table the server wouldquery the pipeline table whether the SMIs is serving theclient If No the SMIs would be loaded in memory anda task-pipeline between the user and SMIs would be builtThereafter the state of the pipeline table would be changed toldquoYesrdquoWhile being Yes the client would be permitted tomakethe next interactive request After that for each request fromthe client the servermust query the state of the behavior tablewhether the pipeline is ldquopendingrdquo or ldquoprocessingrdquo If No theserver would accept the request with a record of a ldquopendingrdquostate Additionally the listener on the server would real-timearrange these pending requests to the server rendering inqueue While a request is being rendered the correspondingstate of behavior table would be changed as ldquoprocessingrdquo toprevent malignant submission While the job is finished thestate would be changed to ldquofinishedrdquo and a 2D projectionimage would be generated and saved in the specified HTTPsharing path
The client adopted AJAX (Asynchronous JavaScript andXML) [31] to dynamically query the state of behavior tablefrom the server at a fixed frequency of 300ms While render-ing was finished the client would download and update the2D projection image via the HTML + js web page technique
3 System Description
31 System Architecture While Figure 1 depicts the Master-Slave dual-channel interaction mode it also presents theoverall architecture of the final platform Here we adoptedthe browserserver architecture for which the communi-cation with each other was via the HTTP protocol Theserver side consists of the Listener amp Request ManagementServer (in zone B) and Rendering Server (in zone C) Theformer accepts and manages the requests from the client
and guides the Rendering Server to an orderly responseThe latter renders the SMIs and generates a 2D projec-tion image for the client downloading via HTTP At thesame time the client side was designed by the standardof HTML5 [32] We used ltimggt an HTML image tagto display the 2D projection image in the form of pseudo3D and used AJAX + js technology to realize flicker-freepage updates and client RIA (rich Internet application)over diversified web browsers without installation More-over HTML5 includes the WebGL standard and the Slavemodel could run on any HTML5-supporting web browserswithout any web plug-ins or special software tools Thusthe whole platform system could run under mature networktechnologies
32 System Front-End System front-end is anHTML5-basedpure website that provides entrance to SMI access and 3Dpresentation The users require a username and password toenter the system or any experts could be invited to enter aspecific task-pipeline with one URL hyperlink from a helpseeker In brief using an HTML5-supporting web browserincluding the major versions of Firefox chrome Operasafari IE 10+ the 3D volume can be viewed and processedas in Figure 6 and the Supplementary Materials availableonline at httpsdoiorg10115520174074137 (1 Demon-stration of the operation procedure on laptop via websitemp4)
Here the interactive behaviors of the user include theviewing angle clipping and coloropacity transformationThefirst two items were operated on the Slave model which wasreconstructed on the client side While the user wants toobserve the final required image of high quality the userjust presses the submit button to submit the current behaviorrequest instruction to the server and waits for the final 2Dprojection image to be updated
Figure 6(a) presents the initial state of the Slave model(right) and 2D projection image (left) In Figure 6(b) mouseor hand gesture is used to operate the Slave model to
International Journal of Telemedicine and Applications 7
Table1Re
lationald
atas
torage
structureo
frequestbehaviorsfrom
clientand
resource
allocatio
non
thes
erver
Table
Field
Note
Basetable
DIC
OMs_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
grou
pof
SMIs
Torecord
theb
asicinform
ationof
SMIs
onthes
erverside
Storagep
ath
Thes
torage
path
ofSM
Ison
thes
erver
side
Pipelin
etable
pipelin
e_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
pipelin
efor
view
ingag
roup
ofSM
Isbetweenthec
lient
andserver
Basedon
basetableto
record
the
conn
ectedrelationshipbetweentheS
MIs
rend
eringin
mem
oryon
thes
ervera
ndremoteu
sero
nthec
lient
DIC
OMs_ID
FORE
IGNKE
Yof
basetableto
decla
rewhich
grou
pof
SMIsthec
lient
isview
ing
Client
Client
user
IPaddress
IPaddresso
fthe
committer
State
Whether
theS
MIsfin
ished
rend
eringin
mem
oryon
thes
erverYesallowthe
clienttonext
operation
Noprompt
for
waitin
g
Behavior
table
behavior_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofeach
behavior
requ
estevent
from
the
client
Basedon
pipelin
etable
torecord
operationrequ
estsfro
mthec
lient
user
andtheirp
rocesssta
te
pipelin
e_ID
FORE
IGNKE
Yof
pipelin
etable
todecla
rewhich
pipelin
eonview
ingand
processin
g
Behavior
type
Inclu
ding
view
inga
ngle
clippingor
coloropacity
transfo
rmation
Behavior-in
structio
n
From
thec
lienttotransfe
rthe
parametersinterm
softhe
correspo
nding
behavior
typerecordedin
theform
ofqu
antizationparameters
State
Pend
ingprocessin
gor
finish
ed
8 International Journal of Telemedicine and Applications
(a) Initial state (b) Viewing angle interaction
(c) Clipping interaction (d) Zooming and coloropacity interaction
Figure 6 Screenshot and operational demonstration
transform the viewing angle from Figure 6(a) to Figure 6(b)(red solid box) and then the submit button is pressed to sendthe behavior request instruction to the server to require anupdated projection image from the Master volume (yellowsolid box) In Figure 6(c) using virtual control panel (reddashed box) to operate Clipping plane after submitting thenew projection image was then updated (left) In Figure 6(d)the mouse wheel or hand gesture is used to zoom the Slavemodel and then the coloropacity theme is submitted via thedropdown list (bottom right corner) to guide remote volumerendering
4 Mode of Availability of the System
In this paper we designed aMaster-Slave dual-channel inter-action mode to improve a problem of remote visualizationinteraction of the volume reconstruction of SMIs whichusually have a huge data size and calculation cost Finallya visualization interaction platform has been built Particu-larly there is no special software and hardware requirementfor the clients and no special network environment isneeded
Tomeasure whether the effectiveness and performance toadapt to the mobile Internet application has characteristicsof lightweight three radiologists with more than 3 years ofexperience were invited to test the interactive experienceimaging quality and compatibility over diversified Internetdevices Moreover four junior students majoring in Biomed-ical Engineeringwere invited to test and quantize the networkload and response time
TestingData 3 groups of SMIs from the radiology departmentof Southwest Hospital of Chongqing include the following
(1) head and neck CT with 495 slices 512 lowast 512 2494MB(2) trunk CT with 609 slices 512 lowast 512 3065MB and (3)thorax andmandible CT with 500 slices 512lowast512 2516MBServer An IBM X3650M4 workstation (CPU 4 times E5-2603RAM 16GB OS Windows Server 2008) was connectedto the Internet via 10Mbps The 2D projection image wasset as the resolution of 512 lowast 512 with high quality JPEGcompression (level-8)
Client 4 Internet devices were located at the 3G or WIFImobile Internet with the server In Figure 7(a) there is aFounder E520 personal computer running on winXP withFirefox 46 web browser via 3G usb card In Figure 7(b) thereis an Acer E15 laptop running on win7 with chrome 43 webbrowser at WIFI In Figure 7(c) there is a HuaWei Honor6 smart phone running on EMUI 30 which is developedfrom Android 442 with the Opera 37 web browser at 3Gnet In Figure 7(d) there is an iPad mini running on IOS6with safari at WIFI All of them could be representative inthe mobile Internet Additionally each was limited to a 100-KBps wireless bandwidth by 360 security firewall for extremetest [33]
41 Interactive Experience Imaging Quality and Compatibil-ity To ensure the consistency of the context the testing data(1) were considered as the main object of reference Here thedemonstration and effect of the testing data (1) are shown inFigure 6
Figure 6 shows that this Slave model of the testing data(1) which was compressed to 104MB could be fit to thedemands of interactive navigation to accurately guide remotevolume rendering and shows that the projection image couldbe displayed to keep the original quality
International Journal of Telemedicine and Applications 9
(a) Personal computer (b) Laptop computer
(c) Smartphone (d) iPad
Figure 7 The system is accessed via four different devices and web browsers under extreme bandwidth
For further validation three radiologists were invited totest the platform with the similar behaviors in Figure 6 Thisjob was repeated to operate three groups of different testingdata through different web browsers and Internet devicesthat were all aforementioned Additionally a five-pointscale was designed to respond to the radiologistsrsquo requests(Table 2)
Table 2 presents a very positive evaluation from the statis-tical result of the imaging quality and interactive experienceand all three radiologists expressed that is an interestingway whether as a gesture-based or mouse-based interactionNo installations were required from the client and goodcompatibility was obtained (same result is shown in Table 3)However one radiologist thought the default Slave modelof the testing data (1) with a skeleton model of 104MB(Figure 3(d2)) was a little rough He suggested to use theskeleton model of 526MB or skin model of 913 KB (Figures3(c1) and 3(c2)) instead of the 104 model both of which arethe alternatives from the mutual interface for the client asdetailed in Section 231 Additionally all three radiologistsclaimed that the response time for final high quality image isalways about 1 second that is consistent with the test resultin Section 42 However they did accept the performancebecause they could directly operate the Slavemodel smoothlywithout any delay on client side thus a short wait for finalhigh quality image had little influence Moreover one of theradiologists specifically expressed the native application of3D rendering on desktop PC also needs 1 or 2 secondsrsquo delayfor final image this performance of the solution running onsmart phone is similar to the native application on PC and isvery good
42 Response Time and Network Load between the Clientand Server The response time can be defined as the timeinterval between the user sending a behavior request tothe system and web browser loading the response resultscompletely The network load can be defined as the amountof data transmission of each interactive behavior Four juniorstudents majoring in Biomedical Engineering were invitedto simultaneously test the platform for one hour During thetest all of the students respectively viewed and processed thetesting data (1) through the platform and performed similarbehaviors to those detailed in Figure 6This test was repeatedby each student with different web browsers and Internetdevices all of which were aforementioned Here we trainedthe students to record the real-time network traffic by the 360security firewall and the time interval by the script log Theresults are showed in Table 3
In Table 3 for each device the amount of data trans-mission and response time of the first loading are muchhigher than the behavior request and idle status because thefirst loading asks the client to download the necessary Slavemodel The response time of the first loading is lt12 s at 100-KBps client bandwidth which is similar to the total timespent on Slavemodel downloading In fact the response timeof first loading also includes the cost time of SMI loadingand rendering on the server side but the server monitorpresents the cost time as lt7 s because SMI loading and Slavemodel downloading are simultaneous and this cost could beneglected
After the first loading the network load of the behaviorrequest for final image is spent on request instructions sub-mission final projection image downloading and the serverprocessing status querying at a fixed frequency of 300ms
10 International Journal of Telemedicine and Applications
Table 2 Statistic results of the five-point scale concerning the interactive experience and imaging quality
Focus Subject Votes from five-point scaleVeryagree Agree Uncertainty Disagree Very
disagree
Interactive experienceMouse-basedGesture-based 998771
The operation of viewing and processing isaccurate
3998771 3
Simple and easy to use without training 2 1998771 3
A good user experience similar to nativeapplication
2 1998771 2 1
Imaging quality
Slavemodel could meet demands of navigation 2 12D image fromMaster volume could provide highquality reconstruction result via utilizing thedisplay capability of Internet terminals
3
Others Latency time for final high quality images 3
Table 3 Average response time and amount of data transmission from the different behaviors and devices at 100-KBps
Devices
First loading (averagevalue)
Behavior request for finalimage (average value)
Idle status ofbandwidthoccupation(KBps)
Frames PerSecond (FPS)Amount of
data trans (KB)Response time(seconds)
Amount ofdata trans (KB)
Response time(seconds)
PC (3G net) 108544 1123 4710 112 023 48ndash60Laptop (WIFI) 106496 1051 4608 089 031 48ndash60iPad mini (WIFI) 107941 1093 4403 095 035 50ndash60Smartphone (3G net) 107502 1072 3994 102 028 30ndash40
where the projection image downloading is the main partThe results present a complete amount of data transmission ofeach behavior request is close to 50KB but the correspondingresponse time for final image is approximately 1 s in 100KBpsThat is because besides the network load the rest of timeis spent on server processing and time interval of querying(300ms) The short delay could be acceptable for practicalapplications as detailed in Section 41
While in the idle status including the userrsquos interactivebehavior (operation) on the Slave model the bandwidthoccupation is almost negligible Moreover there are no sig-nificant differences among diversified devices and networksat 100-KBps client bandwidth
Additionally the FPS of Slave model rendering on theclient side is measured via statsminjs which is a librarypackaged from WebGL When the client rendering is com-pleted FPS could remain at the upper limit
5 Discussion and Comparison
Due to the nonprivate network environment and diversifiedpurposes application software on one device must competefor limited resources of the network and computing Thuslightweight which requires lower resources of the networkand computing and avoids unnecessary installation could bean important characteristic of mobile Internet applicationAdditionally to avoid compatibility problems caused by dif-ferent operating systems and client hardware characteristics
using HTML5 and related web technologies running onmajor web browsers may be a choice to adapt to cross-platform deployment and the characteristics of lightweight[21]
Therefore this article described research on a methodfor the Master-Slave dual-channel interaction mode whichis based on the HTML5 standard (also a wider web com-patibility standard of WebGL) to permit users viewing andprocessing million-megabyte-class SMIs via different clientdevices (without hardware or operating system constraints)over a pure web page (without installing special software) andin a diversified network environment
Compared with the traditional PACS mode of ldquoimagecompression transmission local reconstructionrdquo becauseour method does not transfer original SMIs to the clientit neither needs to wait for a long time to download norrequests special demands of the software environment onthe client device Moreover compared with the state ofthe art of local volume rendering based on WebGL andHTML5 technology [7ndash9] our methods have no specialdemands of client hardware for computation that may reducebattery consumption [9] and enhance the compatibility ofthe client devices our methods may be more fit for mobileInternet application Additionally the client does not accessthe original SMIs which could be much safer
Compared with popular solutions based on absolutesurface rendering technology [11 12 14] our method haslocally used this technique to only support light Slavemodel
International Journal of Telemedicine and Applications 11
for interactive navigation therefore the demands of thenetwork bandwidth and device rendering capability arealmost negligible in ourmethodMoreover the final requiredimage of high quality is a projection drawing from volumerendering technology indicating that our method could beconvenient for users to view the internal anatomy structureand complex organization relationship however surfacerendering technology could not support these characteristicsof volume
Compared with solutions based on absolute remote vol-ume rendering methods our method adopted the Master-Slave dual-channel interaction mode to strengthen the userexperience which is an entirely different approach from themode of ldquocontinuous request continuous responserdquo of studiesof [9] Although studies of [17ndash19] proposed several methodsto improve the mode to reduce the network load includingvideo-compressed transmission [17] tile-based transmission[18] and variable resolution transmission [19] the pressureof server load and real-time data stream increased We usedthe same SMI data (testing data (1)) the same quality offinal required image and the same server configurationand volume rendering algorithm in Section 4 to test themethods in [17ndash19] For quantitative test we formulated aninteractive trajectory with fixed 30 request points theoreti-cally assuming 30 rendering requests triggered per 1 secondthere should be 30 fps render effects On this basis [17]adopted 512 lowast 512 mpeg-4 compression for transmission[18] adopted 8 lowast 8 blocks to segment each 512 lowast 512 frameand transfer the blocks which are different from previousframe in the process of interaction and [19] adopted 64 lowast64 resolution transmission in the process of interaction but512 lowast 512 for end frame The test was repeated for 5 timesunder the condition of one single client monopolizing thebandwidth and server under ideal WIFI environment Theresults showed that the total amount of data transmissionfor each interaction test of [17ndash19] was 2980 KB 6036 KBand 3713 KB respectively the average response time of firstframe was 110 seconds 059 seconds and 052 seconds andthe average cumulative response time of end frame (30thframe which is the final required image) was 2131 seconds1761 seconds and 1601 seconds The response time containsthe factors of server continuous rendering response servercontinuous frame-encoding and so forth (more than 70server CPU utilization during interaction) but the networkfactors could be tentatively neglected under the ideal single-user WIFI environment Assuming that we can improvethe serverrsquos processing ability to completely eliminate theserver response delay (perfect 30 fps interaction effect) thetheoretical value of the bandwidth could be 2980 KBps6036 KBps and 3713 KBps respectively In summary theFPS value one of the most important parameters of the 3Dinteractive experience of the studies of [17ndash19]was correlatedreciprocally with the number of simultaneous tasks on theserver side (server task-pipelines) server processing abilityand network bandwidth and an obvious delay of interactionwas clearly presented By contrast all of the Internet deviceschosen to test our method could provide an FPS value at30ndash60 and the FPS valuewas not correlatedwith server tasksserver processing ability and network bandwidth but with
the device GPU (Graphics Processing Unit) Undoubtedlyalmost all GPUs of Internet devices in the market could meetthe demands of our light Slave model rendering Moreoverthe peak value of the bandwidth under the behavior requestfor final image was less than 50KBps and less than 1 KBpsat idle (including the operation of Slave model) thus bothwere far below those tests of methods of [17ndash19] Thusour method could be more suitable for a complex mobileInternet environment Moreover the complex compressiontransmission methods of [17ndash19] were built on customizedsoftware of plug-ins for the client which is hard to applyto pure web pages and falls short of wide compatibility Bycontrast that is our methodrsquos advantage
For the server load because the Master volume in ourmethod only responds to the final behavior request (instruc-tions) instead of the continuous request of the studies of[9 17ndash19] the test showed that server CPUutilizationwas lessthan 20 at idle (including the operation of Slavemodel) andless than 30 for responding to behavior request thus it wasfar below of those tests ofmethods of [17ndash19] with nearly a fullload during all the operation time under the same conditionsof SMI data quality of final required image server config-uration and volume rendering algorithm Theoretically ourplatform can respond to more independent tasks (pipelines)at the same time Table 2 shows that while the server wasrunning four independent tasks simultaneously the responsetime of our method remained at approximately 1 s in the 100-KBps client bandwidth much better than [17ndash19] which waswith more than 16 secondsrsquo delay of end frame in the singletask status In fact because user viewing and processing ofthe 3D medical volume occurred via interaction on the Slavemodel the delay experience wasmuch fewer than that in [17ndash19] as detailed in Section 41That indicates ourmethod couldhave some more advantages including much less networkload and server load to serve much more independent tasksand users
In addition [17ndash19] did not propose an effective inter-active method to clip the 3D volume what our method hasaccomplished
6 Conclusion
This paper proposed aMaster-Slave dual-channel interactionmode which absorbed the advantages of the surface ren-dering of WebGL and remote volume rendering to builda visualization interaction platform to rapidly view andprocess 3Dmedical volume reconstruction in a pure webpageenvironment This platform could be used in a diversifiednetwork environment (100-KBps bandwidth for extreme)and diversified client devices (no special hardware and oper-ating system constraints) without any special installation butwith a good user experience similar to native applicationThese features could serve authorized users to convenientlyaccess SMIs and present 3D visualization anywhere andcould be a technological basis of online communicationamong doctors and patients [34] Or at least it could helpclinicians to arbitrarily acquire 3D image according to theirclinical needs instead of the static pictures that radiologistsdrafted For example clinicians could conveniently observe
12 International Journal of Telemedicine and Applications
the complicated relationship of skull base bone and bloodvessel via this platform (Figure 6(d)) in the past this pieceof information was provided by a few static images generatedfrom radiologistrsquos understanding
Additionally this solution is designed to be followed asanHTML5 standard interface indicating any authorized usercould rapidly access the application via one URL hyperlinkand the platform could be easily integrated into othertelemedical systems seamlessly as a third-party applicationfor example the Supplementary Materials (2 Demonstrationof the operation procedure on laptop linked by emailmp4and 3 Demonstration of the operation procedure on iPadvia websitemp4) These features may play an assistant roleto improve work patterns and extend the range of applicationof traditional medical image systems For example it can beapplied to regional health telemedicine andmedical imagingeducation at a low cost because there is no special cost ofdeploy and upgrade of software network and hardware onthe client side And more importantly a teacher can vividlyshow the students the image characteristics of the diseaseanywhere a clinician can intuitively introduce disease to thepatients anywhere and ask expertsrsquo help anywhere That maypromote the development of mobile medical technology to acertain degree
Conflicts of Interest
There are no conflicts of interest declared
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China (no 81171866) the National Key BasicResearch Program of China (no 2014CB541602) and theSocial Livelihood Science andTechnology Innovation SpecialProject of CSTC (no cstc2015shmszx120002)
References
[1] Q Zhang M Alexander and L Ryner ldquoSynchronized 2D3Doptical mapping for interactive exploration and real-time visu-alization of multi-function neurological imagesrdquo ComputerizedMedical Imaging and Graphics vol 37 no 7-8 pp 552ndash567 2013
[2] S Zimeras and L G Gortzis ldquoInteractive tele-radiological seg-mentation systems for treatment and diagnosisrdquo InternationalJournal of Telemedicine and Applications vol 2012 Article ID713739 p 15 2012
[3] S E Mahmoudi A Akhondi-Asl R Rahmani et al ldquoWeb-based interactive 2D3D medical image processing andvisualization softwarerdquo Computer Methods and Programs inBiomedicine vol 98 no 2 pp 172ndash182 2010
[4] L Faggioni E Neri I Bargellini et al ldquoIPad-based primary2D reading of CT angiography examinations of patients withsuspected acute gastrointestinal bleeding preliminary experi-ence 1Lrdquo British Journal of Radiology vol 88 no 1047 ArticleID 20140477 2015
[5] J B Park H J Choi J H Lee and B S Kang ldquoAn assessment ofthe iPad 2 as a CT teleradiology tool using brain CT with sub-tle intracranial hemorrhage under conventional illuminationrdquoJournal of Digital Imaging vol 26 no 4 pp 683ndash690 2013
[6] P T Johnson S L Zimmerman D Heath et al ldquoThe iPad asa mobile device for CT display and interpretation diagnosticaccuracy for identification of pulmonary embolismrdquo EmergencyRadiology vol 19 no 4 pp 323ndash327 2012
[7] J M Noguera and J R Jimenez ldquoVisualization of very large 3Dvolumes on mobile devices and WebGLrdquo in Proceedings of the20th International Conference in Central Europe on ComputerGraphics Visualization and Computer Vision WSCG 2012 pp105ndash112 June 2012
[8] J Congote A Segura L Kabongo et al ldquoInteractive visu-alization of volumetric data with WebGL in real-timerdquo inProceedings of the 16th International Conference on 3D WebTechnology (Web3D rsquo11) pp 137ndash145 June 2011
[9] A Schiewe M Anstoots and J Kruger ldquoState of the art inmobile volume rendering on iOS devicesrdquo in Proceedings of theEurographics Conference on Visualization (EuroVis rsquo15) 2015
[10] A Evans M Romeo A Bahrehmand et al ldquo3D graphics on theweb a surveyrdquo Computers amp Graphics vol 41 no 1 pp 43ndash612014
[11] M Callieri R M Andrei M D Benedetto et al ldquoVisualizationmethods formolecular studies on theweb platformrdquo inProceed-ings of the 15th international conference on web 3D technology(Web3D rsquo10) pp 117ndash126 July 2010
[12] J Jimenez A M Lopez J Cruz et al ldquoA Web platform forthe interactive visualization and analysis of the 3D fractaldimension of MRI datardquo Journal of Biomedical Informatics vol51 pp 176ndash190 2014
[13] WebGL httpswwwkhronosorgwebgl[14] C R Butson G Tamm S Jain T Fogal and J Kruger
ldquoEvaluation of interactive visualization on mobile computingplatforms for selection of deep brain stimulation parametersrdquoIEEE Transactions on Visualization and Computer Graphics vol19 no 1 pp 108ndash117 2013
[15] M Levoy ldquoDisplay of surfaces from volume datardquo IEEE Com-puter Graphics and Applications vol 8 no 3 pp 29ndash37 1988
[16] E Kotter T Baumann D Jager and M Langer ldquoTechnologiesfor image distribution in hospitalsrdquo European Radiology vol 16no 6 pp 1270ndash1279 2006
[17] S Park W Kim and I Ihm ldquoMobile collaborative medical dis-play systemrdquo Computer Methods and Programs in Biomedicinevol 89 no 3 pp 248ndash260 2008
[18] J R Mitchell P Sharma J Modi et al ldquoA smartphone client-server teleradiology system for primary diagnosis of acutestrokerdquo Journal of Medical Internet Research vol 13 no 2 p e312011
[19] T Hachaj ldquoReal time exploration and management of largemedical volumetric datasets on small mobile devicesmdasheval-uation of remote volume rendering approachrdquo InternationalJournal of Information Management vol 34 no 3 pp 336ndash3432014
[20] P Quax J Liesenborgs A Barzan et al ldquoRemote renderingsolutions using web technologiesrdquoMultimedia Tools and Appli-cations vol 75 no 8 pp 4383ndash4410 2016
[21] D Preuveneers Y Berbers and W Joosen ldquoThe future ofmobile E-health application development exploring HTML5for context-aware diabetes monitoringrdquo Procedia ComputerScience vol 21 pp 351ndash359 2013
[22] J M Noguera J J Jimenez and M C Osuna-Perez ldquoDevel-opment and evaluation of a 3D mobile application for learningmanual therapy in the physiotherapy laboratoryrdquo Computersand Education vol 69 pp 96ndash108 2013
International Journal of Telemedicine and Applications 13
[23] Y S Kang ldquoVolume rendering overviewrdquo inGPU Programmingand Cg Language Primer chapter 14 pp 154-155 1st edition2009
[24] Visualization Toolkit (VTK) httpwwwvtkorg[25] vtkPiecewiseFunction Class Reference of VTK httpwwwvtk
orgdocnightlyhtmlclassvtkPiecewiseFunctionhtml[26] vtkColorTransferFunction Class Reference of VTK http
wwwvtkorgdocnightlyhtmlannotatedhtml[27] vtkPlane Class Reference of VTK httpwwwvtkorgdoc
nightlyhtmlclassvtkPlanehtml[28] L Ma D-X Zhao and Z-Z Yang ldquoA software tool for
visualization of molecular face (VMF) by improving marchingcubes algorithmrdquo Computational and Theoretical Chemistryvol 1028 pp 34ndash45 2014
[29] G F Wang J Liu and W Liu ldquoDIB display and window trans-formation technology for DICOM medical imagesrdquo ChineseJournal of Medical Device vol 18 no 8 pp 1ndash5 2005
[30] L Qiao X Chen L X Yang et al ldquoDevelopment of medi-cal image three-dimensional-visualization platform based onhome network and pure web conditionsrdquo Beijing BiomedicalEngineering vol 34 no 3 pp 229-233286 2015
[31] M Ying and J Miller ldquoRefactoring legacy AJAX applications toimprove the efficiency of the data exchange componentrdquo Journalof Systems and Software vol 86 no 1 pp 72ndash88 2013
[32] HTML5 httpwwww3orgTR2012CR-html5-20121217[33] QIHU 360 SOFTWARE CO LIMITED 360 Internet Security
httpwww360safecomInternet-securityhtml[34] L Qiao Y Li X Chen et al ldquoMedical high-resolution image
sharing and electronic whiteboard system a pure-web-basedsystem for accessing and discussing lossless original images intelemedicinerdquoComputerMethods and Programs in Biomedicinevol 121 no 2 pp 77ndash91 2015
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
6 International Journal of Telemedicine and Applications
Interactiveeect
Clippingplane
XY
Z
PX
Y
Z
(a) Initial state of Clipping plane
PX
Y
Z
(i) Rotated 30 degrees (ii) Further rotated minus15 degreesaround the X-axis
(ii) Further panning 95
P
XY
Z
P
XY
Z
Y-axisaround the Z-axisalong the
(b) Instances of interaction of Clipping plane (c) Control panelof Clipping plane
Figure 5 Design of the interaction of clipping to the Slavemodel and corresponding example
allocate the system resources for guiding 3D rendering Thecore tables and their data storage structure are designed inTable 1
According to Table 1 if a user chooses a group of SMIsfrom a list derived from the base table the server wouldquery the pipeline table whether the SMIs is serving theclient If No the SMIs would be loaded in memory anda task-pipeline between the user and SMIs would be builtThereafter the state of the pipeline table would be changed toldquoYesrdquoWhile being Yes the client would be permitted tomakethe next interactive request After that for each request fromthe client the servermust query the state of the behavior tablewhether the pipeline is ldquopendingrdquo or ldquoprocessingrdquo If No theserver would accept the request with a record of a ldquopendingrdquostate Additionally the listener on the server would real-timearrange these pending requests to the server rendering inqueue While a request is being rendered the correspondingstate of behavior table would be changed as ldquoprocessingrdquo toprevent malignant submission While the job is finished thestate would be changed to ldquofinishedrdquo and a 2D projectionimage would be generated and saved in the specified HTTPsharing path
The client adopted AJAX (Asynchronous JavaScript andXML) [31] to dynamically query the state of behavior tablefrom the server at a fixed frequency of 300ms While render-ing was finished the client would download and update the2D projection image via the HTML + js web page technique
3 System Description
31 System Architecture While Figure 1 depicts the Master-Slave dual-channel interaction mode it also presents theoverall architecture of the final platform Here we adoptedthe browserserver architecture for which the communi-cation with each other was via the HTTP protocol Theserver side consists of the Listener amp Request ManagementServer (in zone B) and Rendering Server (in zone C) Theformer accepts and manages the requests from the client
and guides the Rendering Server to an orderly responseThe latter renders the SMIs and generates a 2D projec-tion image for the client downloading via HTTP At thesame time the client side was designed by the standardof HTML5 [32] We used ltimggt an HTML image tagto display the 2D projection image in the form of pseudo3D and used AJAX + js technology to realize flicker-freepage updates and client RIA (rich Internet application)over diversified web browsers without installation More-over HTML5 includes the WebGL standard and the Slavemodel could run on any HTML5-supporting web browserswithout any web plug-ins or special software tools Thusthe whole platform system could run under mature networktechnologies
32 System Front-End System front-end is anHTML5-basedpure website that provides entrance to SMI access and 3Dpresentation The users require a username and password toenter the system or any experts could be invited to enter aspecific task-pipeline with one URL hyperlink from a helpseeker In brief using an HTML5-supporting web browserincluding the major versions of Firefox chrome Operasafari IE 10+ the 3D volume can be viewed and processedas in Figure 6 and the Supplementary Materials availableonline at httpsdoiorg10115520174074137 (1 Demon-stration of the operation procedure on laptop via websitemp4)
Here the interactive behaviors of the user include theviewing angle clipping and coloropacity transformationThefirst two items were operated on the Slave model which wasreconstructed on the client side While the user wants toobserve the final required image of high quality the userjust presses the submit button to submit the current behaviorrequest instruction to the server and waits for the final 2Dprojection image to be updated
Figure 6(a) presents the initial state of the Slave model(right) and 2D projection image (left) In Figure 6(b) mouseor hand gesture is used to operate the Slave model to
International Journal of Telemedicine and Applications 7
Table1Re
lationald
atas
torage
structureo
frequestbehaviorsfrom
clientand
resource
allocatio
non
thes
erver
Table
Field
Note
Basetable
DIC
OMs_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
grou
pof
SMIs
Torecord
theb
asicinform
ationof
SMIs
onthes
erverside
Storagep
ath
Thes
torage
path
ofSM
Ison
thes
erver
side
Pipelin
etable
pipelin
e_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
pipelin
efor
view
ingag
roup
ofSM
Isbetweenthec
lient
andserver
Basedon
basetableto
record
the
conn
ectedrelationshipbetweentheS
MIs
rend
eringin
mem
oryon
thes
ervera
ndremoteu
sero
nthec
lient
DIC
OMs_ID
FORE
IGNKE
Yof
basetableto
decla
rewhich
grou
pof
SMIsthec
lient
isview
ing
Client
Client
user
IPaddress
IPaddresso
fthe
committer
State
Whether
theS
MIsfin
ished
rend
eringin
mem
oryon
thes
erverYesallowthe
clienttonext
operation
Noprompt
for
waitin
g
Behavior
table
behavior_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofeach
behavior
requ
estevent
from
the
client
Basedon
pipelin
etable
torecord
operationrequ
estsfro
mthec
lient
user
andtheirp
rocesssta
te
pipelin
e_ID
FORE
IGNKE
Yof
pipelin
etable
todecla
rewhich
pipelin
eonview
ingand
processin
g
Behavior
type
Inclu
ding
view
inga
ngle
clippingor
coloropacity
transfo
rmation
Behavior-in
structio
n
From
thec
lienttotransfe
rthe
parametersinterm
softhe
correspo
nding
behavior
typerecordedin
theform
ofqu
antizationparameters
State
Pend
ingprocessin
gor
finish
ed
8 International Journal of Telemedicine and Applications
(a) Initial state (b) Viewing angle interaction
(c) Clipping interaction (d) Zooming and coloropacity interaction
Figure 6 Screenshot and operational demonstration
transform the viewing angle from Figure 6(a) to Figure 6(b)(red solid box) and then the submit button is pressed to sendthe behavior request instruction to the server to require anupdated projection image from the Master volume (yellowsolid box) In Figure 6(c) using virtual control panel (reddashed box) to operate Clipping plane after submitting thenew projection image was then updated (left) In Figure 6(d)the mouse wheel or hand gesture is used to zoom the Slavemodel and then the coloropacity theme is submitted via thedropdown list (bottom right corner) to guide remote volumerendering
4 Mode of Availability of the System
In this paper we designed aMaster-Slave dual-channel inter-action mode to improve a problem of remote visualizationinteraction of the volume reconstruction of SMIs whichusually have a huge data size and calculation cost Finallya visualization interaction platform has been built Particu-larly there is no special software and hardware requirementfor the clients and no special network environment isneeded
Tomeasure whether the effectiveness and performance toadapt to the mobile Internet application has characteristicsof lightweight three radiologists with more than 3 years ofexperience were invited to test the interactive experienceimaging quality and compatibility over diversified Internetdevices Moreover four junior students majoring in Biomed-ical Engineeringwere invited to test and quantize the networkload and response time
TestingData 3 groups of SMIs from the radiology departmentof Southwest Hospital of Chongqing include the following
(1) head and neck CT with 495 slices 512 lowast 512 2494MB(2) trunk CT with 609 slices 512 lowast 512 3065MB and (3)thorax andmandible CT with 500 slices 512lowast512 2516MBServer An IBM X3650M4 workstation (CPU 4 times E5-2603RAM 16GB OS Windows Server 2008) was connectedto the Internet via 10Mbps The 2D projection image wasset as the resolution of 512 lowast 512 with high quality JPEGcompression (level-8)
Client 4 Internet devices were located at the 3G or WIFImobile Internet with the server In Figure 7(a) there is aFounder E520 personal computer running on winXP withFirefox 46 web browser via 3G usb card In Figure 7(b) thereis an Acer E15 laptop running on win7 with chrome 43 webbrowser at WIFI In Figure 7(c) there is a HuaWei Honor6 smart phone running on EMUI 30 which is developedfrom Android 442 with the Opera 37 web browser at 3Gnet In Figure 7(d) there is an iPad mini running on IOS6with safari at WIFI All of them could be representative inthe mobile Internet Additionally each was limited to a 100-KBps wireless bandwidth by 360 security firewall for extremetest [33]
41 Interactive Experience Imaging Quality and Compatibil-ity To ensure the consistency of the context the testing data(1) were considered as the main object of reference Here thedemonstration and effect of the testing data (1) are shown inFigure 6
Figure 6 shows that this Slave model of the testing data(1) which was compressed to 104MB could be fit to thedemands of interactive navigation to accurately guide remotevolume rendering and shows that the projection image couldbe displayed to keep the original quality
International Journal of Telemedicine and Applications 9
(a) Personal computer (b) Laptop computer
(c) Smartphone (d) iPad
Figure 7 The system is accessed via four different devices and web browsers under extreme bandwidth
For further validation three radiologists were invited totest the platform with the similar behaviors in Figure 6 Thisjob was repeated to operate three groups of different testingdata through different web browsers and Internet devicesthat were all aforementioned Additionally a five-pointscale was designed to respond to the radiologistsrsquo requests(Table 2)
Table 2 presents a very positive evaluation from the statis-tical result of the imaging quality and interactive experienceand all three radiologists expressed that is an interestingway whether as a gesture-based or mouse-based interactionNo installations were required from the client and goodcompatibility was obtained (same result is shown in Table 3)However one radiologist thought the default Slave modelof the testing data (1) with a skeleton model of 104MB(Figure 3(d2)) was a little rough He suggested to use theskeleton model of 526MB or skin model of 913 KB (Figures3(c1) and 3(c2)) instead of the 104 model both of which arethe alternatives from the mutual interface for the client asdetailed in Section 231 Additionally all three radiologistsclaimed that the response time for final high quality image isalways about 1 second that is consistent with the test resultin Section 42 However they did accept the performancebecause they could directly operate the Slavemodel smoothlywithout any delay on client side thus a short wait for finalhigh quality image had little influence Moreover one of theradiologists specifically expressed the native application of3D rendering on desktop PC also needs 1 or 2 secondsrsquo delayfor final image this performance of the solution running onsmart phone is similar to the native application on PC and isvery good
42 Response Time and Network Load between the Clientand Server The response time can be defined as the timeinterval between the user sending a behavior request tothe system and web browser loading the response resultscompletely The network load can be defined as the amountof data transmission of each interactive behavior Four juniorstudents majoring in Biomedical Engineering were invitedto simultaneously test the platform for one hour During thetest all of the students respectively viewed and processed thetesting data (1) through the platform and performed similarbehaviors to those detailed in Figure 6This test was repeatedby each student with different web browsers and Internetdevices all of which were aforementioned Here we trainedthe students to record the real-time network traffic by the 360security firewall and the time interval by the script log Theresults are showed in Table 3
In Table 3 for each device the amount of data trans-mission and response time of the first loading are muchhigher than the behavior request and idle status because thefirst loading asks the client to download the necessary Slavemodel The response time of the first loading is lt12 s at 100-KBps client bandwidth which is similar to the total timespent on Slavemodel downloading In fact the response timeof first loading also includes the cost time of SMI loadingand rendering on the server side but the server monitorpresents the cost time as lt7 s because SMI loading and Slavemodel downloading are simultaneous and this cost could beneglected
After the first loading the network load of the behaviorrequest for final image is spent on request instructions sub-mission final projection image downloading and the serverprocessing status querying at a fixed frequency of 300ms
10 International Journal of Telemedicine and Applications
Table 2 Statistic results of the five-point scale concerning the interactive experience and imaging quality
Focus Subject Votes from five-point scaleVeryagree Agree Uncertainty Disagree Very
disagree
Interactive experienceMouse-basedGesture-based 998771
The operation of viewing and processing isaccurate
3998771 3
Simple and easy to use without training 2 1998771 3
A good user experience similar to nativeapplication
2 1998771 2 1
Imaging quality
Slavemodel could meet demands of navigation 2 12D image fromMaster volume could provide highquality reconstruction result via utilizing thedisplay capability of Internet terminals
3
Others Latency time for final high quality images 3
Table 3 Average response time and amount of data transmission from the different behaviors and devices at 100-KBps
Devices
First loading (averagevalue)
Behavior request for finalimage (average value)
Idle status ofbandwidthoccupation(KBps)
Frames PerSecond (FPS)Amount of
data trans (KB)Response time(seconds)
Amount ofdata trans (KB)
Response time(seconds)
PC (3G net) 108544 1123 4710 112 023 48ndash60Laptop (WIFI) 106496 1051 4608 089 031 48ndash60iPad mini (WIFI) 107941 1093 4403 095 035 50ndash60Smartphone (3G net) 107502 1072 3994 102 028 30ndash40
where the projection image downloading is the main partThe results present a complete amount of data transmission ofeach behavior request is close to 50KB but the correspondingresponse time for final image is approximately 1 s in 100KBpsThat is because besides the network load the rest of timeis spent on server processing and time interval of querying(300ms) The short delay could be acceptable for practicalapplications as detailed in Section 41
While in the idle status including the userrsquos interactivebehavior (operation) on the Slave model the bandwidthoccupation is almost negligible Moreover there are no sig-nificant differences among diversified devices and networksat 100-KBps client bandwidth
Additionally the FPS of Slave model rendering on theclient side is measured via statsminjs which is a librarypackaged from WebGL When the client rendering is com-pleted FPS could remain at the upper limit
5 Discussion and Comparison
Due to the nonprivate network environment and diversifiedpurposes application software on one device must competefor limited resources of the network and computing Thuslightweight which requires lower resources of the networkand computing and avoids unnecessary installation could bean important characteristic of mobile Internet applicationAdditionally to avoid compatibility problems caused by dif-ferent operating systems and client hardware characteristics
using HTML5 and related web technologies running onmajor web browsers may be a choice to adapt to cross-platform deployment and the characteristics of lightweight[21]
Therefore this article described research on a methodfor the Master-Slave dual-channel interaction mode whichis based on the HTML5 standard (also a wider web com-patibility standard of WebGL) to permit users viewing andprocessing million-megabyte-class SMIs via different clientdevices (without hardware or operating system constraints)over a pure web page (without installing special software) andin a diversified network environment
Compared with the traditional PACS mode of ldquoimagecompression transmission local reconstructionrdquo becauseour method does not transfer original SMIs to the clientit neither needs to wait for a long time to download norrequests special demands of the software environment onthe client device Moreover compared with the state ofthe art of local volume rendering based on WebGL andHTML5 technology [7ndash9] our methods have no specialdemands of client hardware for computation that may reducebattery consumption [9] and enhance the compatibility ofthe client devices our methods may be more fit for mobileInternet application Additionally the client does not accessthe original SMIs which could be much safer
Compared with popular solutions based on absolutesurface rendering technology [11 12 14] our method haslocally used this technique to only support light Slavemodel
International Journal of Telemedicine and Applications 11
for interactive navigation therefore the demands of thenetwork bandwidth and device rendering capability arealmost negligible in ourmethodMoreover the final requiredimage of high quality is a projection drawing from volumerendering technology indicating that our method could beconvenient for users to view the internal anatomy structureand complex organization relationship however surfacerendering technology could not support these characteristicsof volume
Compared with solutions based on absolute remote vol-ume rendering methods our method adopted the Master-Slave dual-channel interaction mode to strengthen the userexperience which is an entirely different approach from themode of ldquocontinuous request continuous responserdquo of studiesof [9] Although studies of [17ndash19] proposed several methodsto improve the mode to reduce the network load includingvideo-compressed transmission [17] tile-based transmission[18] and variable resolution transmission [19] the pressureof server load and real-time data stream increased We usedthe same SMI data (testing data (1)) the same quality offinal required image and the same server configurationand volume rendering algorithm in Section 4 to test themethods in [17ndash19] For quantitative test we formulated aninteractive trajectory with fixed 30 request points theoreti-cally assuming 30 rendering requests triggered per 1 secondthere should be 30 fps render effects On this basis [17]adopted 512 lowast 512 mpeg-4 compression for transmission[18] adopted 8 lowast 8 blocks to segment each 512 lowast 512 frameand transfer the blocks which are different from previousframe in the process of interaction and [19] adopted 64 lowast64 resolution transmission in the process of interaction but512 lowast 512 for end frame The test was repeated for 5 timesunder the condition of one single client monopolizing thebandwidth and server under ideal WIFI environment Theresults showed that the total amount of data transmissionfor each interaction test of [17ndash19] was 2980 KB 6036 KBand 3713 KB respectively the average response time of firstframe was 110 seconds 059 seconds and 052 seconds andthe average cumulative response time of end frame (30thframe which is the final required image) was 2131 seconds1761 seconds and 1601 seconds The response time containsthe factors of server continuous rendering response servercontinuous frame-encoding and so forth (more than 70server CPU utilization during interaction) but the networkfactors could be tentatively neglected under the ideal single-user WIFI environment Assuming that we can improvethe serverrsquos processing ability to completely eliminate theserver response delay (perfect 30 fps interaction effect) thetheoretical value of the bandwidth could be 2980 KBps6036 KBps and 3713 KBps respectively In summary theFPS value one of the most important parameters of the 3Dinteractive experience of the studies of [17ndash19]was correlatedreciprocally with the number of simultaneous tasks on theserver side (server task-pipelines) server processing abilityand network bandwidth and an obvious delay of interactionwas clearly presented By contrast all of the Internet deviceschosen to test our method could provide an FPS value at30ndash60 and the FPS valuewas not correlatedwith server tasksserver processing ability and network bandwidth but with
the device GPU (Graphics Processing Unit) Undoubtedlyalmost all GPUs of Internet devices in the market could meetthe demands of our light Slave model rendering Moreoverthe peak value of the bandwidth under the behavior requestfor final image was less than 50KBps and less than 1 KBpsat idle (including the operation of Slave model) thus bothwere far below those tests of methods of [17ndash19] Thusour method could be more suitable for a complex mobileInternet environment Moreover the complex compressiontransmission methods of [17ndash19] were built on customizedsoftware of plug-ins for the client which is hard to applyto pure web pages and falls short of wide compatibility Bycontrast that is our methodrsquos advantage
For the server load because the Master volume in ourmethod only responds to the final behavior request (instruc-tions) instead of the continuous request of the studies of[9 17ndash19] the test showed that server CPUutilizationwas lessthan 20 at idle (including the operation of Slavemodel) andless than 30 for responding to behavior request thus it wasfar below of those tests ofmethods of [17ndash19] with nearly a fullload during all the operation time under the same conditionsof SMI data quality of final required image server config-uration and volume rendering algorithm Theoretically ourplatform can respond to more independent tasks (pipelines)at the same time Table 2 shows that while the server wasrunning four independent tasks simultaneously the responsetime of our method remained at approximately 1 s in the 100-KBps client bandwidth much better than [17ndash19] which waswith more than 16 secondsrsquo delay of end frame in the singletask status In fact because user viewing and processing ofthe 3D medical volume occurred via interaction on the Slavemodel the delay experience wasmuch fewer than that in [17ndash19] as detailed in Section 41That indicates ourmethod couldhave some more advantages including much less networkload and server load to serve much more independent tasksand users
In addition [17ndash19] did not propose an effective inter-active method to clip the 3D volume what our method hasaccomplished
6 Conclusion
This paper proposed aMaster-Slave dual-channel interactionmode which absorbed the advantages of the surface ren-dering of WebGL and remote volume rendering to builda visualization interaction platform to rapidly view andprocess 3Dmedical volume reconstruction in a pure webpageenvironment This platform could be used in a diversifiednetwork environment (100-KBps bandwidth for extreme)and diversified client devices (no special hardware and oper-ating system constraints) without any special installation butwith a good user experience similar to native applicationThese features could serve authorized users to convenientlyaccess SMIs and present 3D visualization anywhere andcould be a technological basis of online communicationamong doctors and patients [34] Or at least it could helpclinicians to arbitrarily acquire 3D image according to theirclinical needs instead of the static pictures that radiologistsdrafted For example clinicians could conveniently observe
12 International Journal of Telemedicine and Applications
the complicated relationship of skull base bone and bloodvessel via this platform (Figure 6(d)) in the past this pieceof information was provided by a few static images generatedfrom radiologistrsquos understanding
Additionally this solution is designed to be followed asanHTML5 standard interface indicating any authorized usercould rapidly access the application via one URL hyperlinkand the platform could be easily integrated into othertelemedical systems seamlessly as a third-party applicationfor example the Supplementary Materials (2 Demonstrationof the operation procedure on laptop linked by emailmp4and 3 Demonstration of the operation procedure on iPadvia websitemp4) These features may play an assistant roleto improve work patterns and extend the range of applicationof traditional medical image systems For example it can beapplied to regional health telemedicine andmedical imagingeducation at a low cost because there is no special cost ofdeploy and upgrade of software network and hardware onthe client side And more importantly a teacher can vividlyshow the students the image characteristics of the diseaseanywhere a clinician can intuitively introduce disease to thepatients anywhere and ask expertsrsquo help anywhere That maypromote the development of mobile medical technology to acertain degree
Conflicts of Interest
There are no conflicts of interest declared
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China (no 81171866) the National Key BasicResearch Program of China (no 2014CB541602) and theSocial Livelihood Science andTechnology Innovation SpecialProject of CSTC (no cstc2015shmszx120002)
References
[1] Q Zhang M Alexander and L Ryner ldquoSynchronized 2D3Doptical mapping for interactive exploration and real-time visu-alization of multi-function neurological imagesrdquo ComputerizedMedical Imaging and Graphics vol 37 no 7-8 pp 552ndash567 2013
[2] S Zimeras and L G Gortzis ldquoInteractive tele-radiological seg-mentation systems for treatment and diagnosisrdquo InternationalJournal of Telemedicine and Applications vol 2012 Article ID713739 p 15 2012
[3] S E Mahmoudi A Akhondi-Asl R Rahmani et al ldquoWeb-based interactive 2D3D medical image processing andvisualization softwarerdquo Computer Methods and Programs inBiomedicine vol 98 no 2 pp 172ndash182 2010
[4] L Faggioni E Neri I Bargellini et al ldquoIPad-based primary2D reading of CT angiography examinations of patients withsuspected acute gastrointestinal bleeding preliminary experi-ence 1Lrdquo British Journal of Radiology vol 88 no 1047 ArticleID 20140477 2015
[5] J B Park H J Choi J H Lee and B S Kang ldquoAn assessment ofthe iPad 2 as a CT teleradiology tool using brain CT with sub-tle intracranial hemorrhage under conventional illuminationrdquoJournal of Digital Imaging vol 26 no 4 pp 683ndash690 2013
[6] P T Johnson S L Zimmerman D Heath et al ldquoThe iPad asa mobile device for CT display and interpretation diagnosticaccuracy for identification of pulmonary embolismrdquo EmergencyRadiology vol 19 no 4 pp 323ndash327 2012
[7] J M Noguera and J R Jimenez ldquoVisualization of very large 3Dvolumes on mobile devices and WebGLrdquo in Proceedings of the20th International Conference in Central Europe on ComputerGraphics Visualization and Computer Vision WSCG 2012 pp105ndash112 June 2012
[8] J Congote A Segura L Kabongo et al ldquoInteractive visu-alization of volumetric data with WebGL in real-timerdquo inProceedings of the 16th International Conference on 3D WebTechnology (Web3D rsquo11) pp 137ndash145 June 2011
[9] A Schiewe M Anstoots and J Kruger ldquoState of the art inmobile volume rendering on iOS devicesrdquo in Proceedings of theEurographics Conference on Visualization (EuroVis rsquo15) 2015
[10] A Evans M Romeo A Bahrehmand et al ldquo3D graphics on theweb a surveyrdquo Computers amp Graphics vol 41 no 1 pp 43ndash612014
[11] M Callieri R M Andrei M D Benedetto et al ldquoVisualizationmethods formolecular studies on theweb platformrdquo inProceed-ings of the 15th international conference on web 3D technology(Web3D rsquo10) pp 117ndash126 July 2010
[12] J Jimenez A M Lopez J Cruz et al ldquoA Web platform forthe interactive visualization and analysis of the 3D fractaldimension of MRI datardquo Journal of Biomedical Informatics vol51 pp 176ndash190 2014
[13] WebGL httpswwwkhronosorgwebgl[14] C R Butson G Tamm S Jain T Fogal and J Kruger
ldquoEvaluation of interactive visualization on mobile computingplatforms for selection of deep brain stimulation parametersrdquoIEEE Transactions on Visualization and Computer Graphics vol19 no 1 pp 108ndash117 2013
[15] M Levoy ldquoDisplay of surfaces from volume datardquo IEEE Com-puter Graphics and Applications vol 8 no 3 pp 29ndash37 1988
[16] E Kotter T Baumann D Jager and M Langer ldquoTechnologiesfor image distribution in hospitalsrdquo European Radiology vol 16no 6 pp 1270ndash1279 2006
[17] S Park W Kim and I Ihm ldquoMobile collaborative medical dis-play systemrdquo Computer Methods and Programs in Biomedicinevol 89 no 3 pp 248ndash260 2008
[18] J R Mitchell P Sharma J Modi et al ldquoA smartphone client-server teleradiology system for primary diagnosis of acutestrokerdquo Journal of Medical Internet Research vol 13 no 2 p e312011
[19] T Hachaj ldquoReal time exploration and management of largemedical volumetric datasets on small mobile devicesmdasheval-uation of remote volume rendering approachrdquo InternationalJournal of Information Management vol 34 no 3 pp 336ndash3432014
[20] P Quax J Liesenborgs A Barzan et al ldquoRemote renderingsolutions using web technologiesrdquoMultimedia Tools and Appli-cations vol 75 no 8 pp 4383ndash4410 2016
[21] D Preuveneers Y Berbers and W Joosen ldquoThe future ofmobile E-health application development exploring HTML5for context-aware diabetes monitoringrdquo Procedia ComputerScience vol 21 pp 351ndash359 2013
[22] J M Noguera J J Jimenez and M C Osuna-Perez ldquoDevel-opment and evaluation of a 3D mobile application for learningmanual therapy in the physiotherapy laboratoryrdquo Computersand Education vol 69 pp 96ndash108 2013
International Journal of Telemedicine and Applications 13
[23] Y S Kang ldquoVolume rendering overviewrdquo inGPU Programmingand Cg Language Primer chapter 14 pp 154-155 1st edition2009
[24] Visualization Toolkit (VTK) httpwwwvtkorg[25] vtkPiecewiseFunction Class Reference of VTK httpwwwvtk
orgdocnightlyhtmlclassvtkPiecewiseFunctionhtml[26] vtkColorTransferFunction Class Reference of VTK http
wwwvtkorgdocnightlyhtmlannotatedhtml[27] vtkPlane Class Reference of VTK httpwwwvtkorgdoc
nightlyhtmlclassvtkPlanehtml[28] L Ma D-X Zhao and Z-Z Yang ldquoA software tool for
visualization of molecular face (VMF) by improving marchingcubes algorithmrdquo Computational and Theoretical Chemistryvol 1028 pp 34ndash45 2014
[29] G F Wang J Liu and W Liu ldquoDIB display and window trans-formation technology for DICOM medical imagesrdquo ChineseJournal of Medical Device vol 18 no 8 pp 1ndash5 2005
[30] L Qiao X Chen L X Yang et al ldquoDevelopment of medi-cal image three-dimensional-visualization platform based onhome network and pure web conditionsrdquo Beijing BiomedicalEngineering vol 34 no 3 pp 229-233286 2015
[31] M Ying and J Miller ldquoRefactoring legacy AJAX applications toimprove the efficiency of the data exchange componentrdquo Journalof Systems and Software vol 86 no 1 pp 72ndash88 2013
[32] HTML5 httpwwww3orgTR2012CR-html5-20121217[33] QIHU 360 SOFTWARE CO LIMITED 360 Internet Security
httpwww360safecomInternet-securityhtml[34] L Qiao Y Li X Chen et al ldquoMedical high-resolution image
sharing and electronic whiteboard system a pure-web-basedsystem for accessing and discussing lossless original images intelemedicinerdquoComputerMethods and Programs in Biomedicinevol 121 no 2 pp 77ndash91 2015
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of Telemedicine and Applications 7
Table1Re
lationald
atas
torage
structureo
frequestbehaviorsfrom
clientand
resource
allocatio
non
thes
erver
Table
Field
Note
Basetable
DIC
OMs_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
grou
pof
SMIs
Torecord
theb
asicinform
ationof
SMIs
onthes
erverside
Storagep
ath
Thes
torage
path
ofSM
Ison
thes
erver
side
Pipelin
etable
pipelin
e_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofa
pipelin
efor
view
ingag
roup
ofSM
Isbetweenthec
lient
andserver
Basedon
basetableto
record
the
conn
ectedrelationshipbetweentheS
MIs
rend
eringin
mem
oryon
thes
ervera
ndremoteu
sero
nthec
lient
DIC
OMs_ID
FORE
IGNKE
Yof
basetableto
decla
rewhich
grou
pof
SMIsthec
lient
isview
ing
Client
Client
user
IPaddress
IPaddresso
fthe
committer
State
Whether
theS
MIsfin
ished
rend
eringin
mem
oryon
thes
erverYesallowthe
clienttonext
operation
Noprompt
for
waitin
g
Behavior
table
behavior_ID
PRIM
ARY
KEYtheu
niqu
eidentifier
ofeach
behavior
requ
estevent
from
the
client
Basedon
pipelin
etable
torecord
operationrequ
estsfro
mthec
lient
user
andtheirp
rocesssta
te
pipelin
e_ID
FORE
IGNKE
Yof
pipelin
etable
todecla
rewhich
pipelin
eonview
ingand
processin
g
Behavior
type
Inclu
ding
view
inga
ngle
clippingor
coloropacity
transfo
rmation
Behavior-in
structio
n
From
thec
lienttotransfe
rthe
parametersinterm
softhe
correspo
nding
behavior
typerecordedin
theform
ofqu
antizationparameters
State
Pend
ingprocessin
gor
finish
ed
8 International Journal of Telemedicine and Applications
(a) Initial state (b) Viewing angle interaction
(c) Clipping interaction (d) Zooming and coloropacity interaction
Figure 6 Screenshot and operational demonstration
transform the viewing angle from Figure 6(a) to Figure 6(b)(red solid box) and then the submit button is pressed to sendthe behavior request instruction to the server to require anupdated projection image from the Master volume (yellowsolid box) In Figure 6(c) using virtual control panel (reddashed box) to operate Clipping plane after submitting thenew projection image was then updated (left) In Figure 6(d)the mouse wheel or hand gesture is used to zoom the Slavemodel and then the coloropacity theme is submitted via thedropdown list (bottom right corner) to guide remote volumerendering
4 Mode of Availability of the System
In this paper we designed aMaster-Slave dual-channel inter-action mode to improve a problem of remote visualizationinteraction of the volume reconstruction of SMIs whichusually have a huge data size and calculation cost Finallya visualization interaction platform has been built Particu-larly there is no special software and hardware requirementfor the clients and no special network environment isneeded
Tomeasure whether the effectiveness and performance toadapt to the mobile Internet application has characteristicsof lightweight three radiologists with more than 3 years ofexperience were invited to test the interactive experienceimaging quality and compatibility over diversified Internetdevices Moreover four junior students majoring in Biomed-ical Engineeringwere invited to test and quantize the networkload and response time
TestingData 3 groups of SMIs from the radiology departmentof Southwest Hospital of Chongqing include the following
(1) head and neck CT with 495 slices 512 lowast 512 2494MB(2) trunk CT with 609 slices 512 lowast 512 3065MB and (3)thorax andmandible CT with 500 slices 512lowast512 2516MBServer An IBM X3650M4 workstation (CPU 4 times E5-2603RAM 16GB OS Windows Server 2008) was connectedto the Internet via 10Mbps The 2D projection image wasset as the resolution of 512 lowast 512 with high quality JPEGcompression (level-8)
Client 4 Internet devices were located at the 3G or WIFImobile Internet with the server In Figure 7(a) there is aFounder E520 personal computer running on winXP withFirefox 46 web browser via 3G usb card In Figure 7(b) thereis an Acer E15 laptop running on win7 with chrome 43 webbrowser at WIFI In Figure 7(c) there is a HuaWei Honor6 smart phone running on EMUI 30 which is developedfrom Android 442 with the Opera 37 web browser at 3Gnet In Figure 7(d) there is an iPad mini running on IOS6with safari at WIFI All of them could be representative inthe mobile Internet Additionally each was limited to a 100-KBps wireless bandwidth by 360 security firewall for extremetest [33]
41 Interactive Experience Imaging Quality and Compatibil-ity To ensure the consistency of the context the testing data(1) were considered as the main object of reference Here thedemonstration and effect of the testing data (1) are shown inFigure 6
Figure 6 shows that this Slave model of the testing data(1) which was compressed to 104MB could be fit to thedemands of interactive navigation to accurately guide remotevolume rendering and shows that the projection image couldbe displayed to keep the original quality
International Journal of Telemedicine and Applications 9
(a) Personal computer (b) Laptop computer
(c) Smartphone (d) iPad
Figure 7 The system is accessed via four different devices and web browsers under extreme bandwidth
For further validation three radiologists were invited totest the platform with the similar behaviors in Figure 6 Thisjob was repeated to operate three groups of different testingdata through different web browsers and Internet devicesthat were all aforementioned Additionally a five-pointscale was designed to respond to the radiologistsrsquo requests(Table 2)
Table 2 presents a very positive evaluation from the statis-tical result of the imaging quality and interactive experienceand all three radiologists expressed that is an interestingway whether as a gesture-based or mouse-based interactionNo installations were required from the client and goodcompatibility was obtained (same result is shown in Table 3)However one radiologist thought the default Slave modelof the testing data (1) with a skeleton model of 104MB(Figure 3(d2)) was a little rough He suggested to use theskeleton model of 526MB or skin model of 913 KB (Figures3(c1) and 3(c2)) instead of the 104 model both of which arethe alternatives from the mutual interface for the client asdetailed in Section 231 Additionally all three radiologistsclaimed that the response time for final high quality image isalways about 1 second that is consistent with the test resultin Section 42 However they did accept the performancebecause they could directly operate the Slavemodel smoothlywithout any delay on client side thus a short wait for finalhigh quality image had little influence Moreover one of theradiologists specifically expressed the native application of3D rendering on desktop PC also needs 1 or 2 secondsrsquo delayfor final image this performance of the solution running onsmart phone is similar to the native application on PC and isvery good
42 Response Time and Network Load between the Clientand Server The response time can be defined as the timeinterval between the user sending a behavior request tothe system and web browser loading the response resultscompletely The network load can be defined as the amountof data transmission of each interactive behavior Four juniorstudents majoring in Biomedical Engineering were invitedto simultaneously test the platform for one hour During thetest all of the students respectively viewed and processed thetesting data (1) through the platform and performed similarbehaviors to those detailed in Figure 6This test was repeatedby each student with different web browsers and Internetdevices all of which were aforementioned Here we trainedthe students to record the real-time network traffic by the 360security firewall and the time interval by the script log Theresults are showed in Table 3
In Table 3 for each device the amount of data trans-mission and response time of the first loading are muchhigher than the behavior request and idle status because thefirst loading asks the client to download the necessary Slavemodel The response time of the first loading is lt12 s at 100-KBps client bandwidth which is similar to the total timespent on Slavemodel downloading In fact the response timeof first loading also includes the cost time of SMI loadingand rendering on the server side but the server monitorpresents the cost time as lt7 s because SMI loading and Slavemodel downloading are simultaneous and this cost could beneglected
After the first loading the network load of the behaviorrequest for final image is spent on request instructions sub-mission final projection image downloading and the serverprocessing status querying at a fixed frequency of 300ms
10 International Journal of Telemedicine and Applications
Table 2 Statistic results of the five-point scale concerning the interactive experience and imaging quality
Focus Subject Votes from five-point scaleVeryagree Agree Uncertainty Disagree Very
disagree
Interactive experienceMouse-basedGesture-based 998771
The operation of viewing and processing isaccurate
3998771 3
Simple and easy to use without training 2 1998771 3
A good user experience similar to nativeapplication
2 1998771 2 1
Imaging quality
Slavemodel could meet demands of navigation 2 12D image fromMaster volume could provide highquality reconstruction result via utilizing thedisplay capability of Internet terminals
3
Others Latency time for final high quality images 3
Table 3 Average response time and amount of data transmission from the different behaviors and devices at 100-KBps
Devices
First loading (averagevalue)
Behavior request for finalimage (average value)
Idle status ofbandwidthoccupation(KBps)
Frames PerSecond (FPS)Amount of
data trans (KB)Response time(seconds)
Amount ofdata trans (KB)
Response time(seconds)
PC (3G net) 108544 1123 4710 112 023 48ndash60Laptop (WIFI) 106496 1051 4608 089 031 48ndash60iPad mini (WIFI) 107941 1093 4403 095 035 50ndash60Smartphone (3G net) 107502 1072 3994 102 028 30ndash40
where the projection image downloading is the main partThe results present a complete amount of data transmission ofeach behavior request is close to 50KB but the correspondingresponse time for final image is approximately 1 s in 100KBpsThat is because besides the network load the rest of timeis spent on server processing and time interval of querying(300ms) The short delay could be acceptable for practicalapplications as detailed in Section 41
While in the idle status including the userrsquos interactivebehavior (operation) on the Slave model the bandwidthoccupation is almost negligible Moreover there are no sig-nificant differences among diversified devices and networksat 100-KBps client bandwidth
Additionally the FPS of Slave model rendering on theclient side is measured via statsminjs which is a librarypackaged from WebGL When the client rendering is com-pleted FPS could remain at the upper limit
5 Discussion and Comparison
Due to the nonprivate network environment and diversifiedpurposes application software on one device must competefor limited resources of the network and computing Thuslightweight which requires lower resources of the networkand computing and avoids unnecessary installation could bean important characteristic of mobile Internet applicationAdditionally to avoid compatibility problems caused by dif-ferent operating systems and client hardware characteristics
using HTML5 and related web technologies running onmajor web browsers may be a choice to adapt to cross-platform deployment and the characteristics of lightweight[21]
Therefore this article described research on a methodfor the Master-Slave dual-channel interaction mode whichis based on the HTML5 standard (also a wider web com-patibility standard of WebGL) to permit users viewing andprocessing million-megabyte-class SMIs via different clientdevices (without hardware or operating system constraints)over a pure web page (without installing special software) andin a diversified network environment
Compared with the traditional PACS mode of ldquoimagecompression transmission local reconstructionrdquo becauseour method does not transfer original SMIs to the clientit neither needs to wait for a long time to download norrequests special demands of the software environment onthe client device Moreover compared with the state ofthe art of local volume rendering based on WebGL andHTML5 technology [7ndash9] our methods have no specialdemands of client hardware for computation that may reducebattery consumption [9] and enhance the compatibility ofthe client devices our methods may be more fit for mobileInternet application Additionally the client does not accessthe original SMIs which could be much safer
Compared with popular solutions based on absolutesurface rendering technology [11 12 14] our method haslocally used this technique to only support light Slavemodel
International Journal of Telemedicine and Applications 11
for interactive navigation therefore the demands of thenetwork bandwidth and device rendering capability arealmost negligible in ourmethodMoreover the final requiredimage of high quality is a projection drawing from volumerendering technology indicating that our method could beconvenient for users to view the internal anatomy structureand complex organization relationship however surfacerendering technology could not support these characteristicsof volume
Compared with solutions based on absolute remote vol-ume rendering methods our method adopted the Master-Slave dual-channel interaction mode to strengthen the userexperience which is an entirely different approach from themode of ldquocontinuous request continuous responserdquo of studiesof [9] Although studies of [17ndash19] proposed several methodsto improve the mode to reduce the network load includingvideo-compressed transmission [17] tile-based transmission[18] and variable resolution transmission [19] the pressureof server load and real-time data stream increased We usedthe same SMI data (testing data (1)) the same quality offinal required image and the same server configurationand volume rendering algorithm in Section 4 to test themethods in [17ndash19] For quantitative test we formulated aninteractive trajectory with fixed 30 request points theoreti-cally assuming 30 rendering requests triggered per 1 secondthere should be 30 fps render effects On this basis [17]adopted 512 lowast 512 mpeg-4 compression for transmission[18] adopted 8 lowast 8 blocks to segment each 512 lowast 512 frameand transfer the blocks which are different from previousframe in the process of interaction and [19] adopted 64 lowast64 resolution transmission in the process of interaction but512 lowast 512 for end frame The test was repeated for 5 timesunder the condition of one single client monopolizing thebandwidth and server under ideal WIFI environment Theresults showed that the total amount of data transmissionfor each interaction test of [17ndash19] was 2980 KB 6036 KBand 3713 KB respectively the average response time of firstframe was 110 seconds 059 seconds and 052 seconds andthe average cumulative response time of end frame (30thframe which is the final required image) was 2131 seconds1761 seconds and 1601 seconds The response time containsthe factors of server continuous rendering response servercontinuous frame-encoding and so forth (more than 70server CPU utilization during interaction) but the networkfactors could be tentatively neglected under the ideal single-user WIFI environment Assuming that we can improvethe serverrsquos processing ability to completely eliminate theserver response delay (perfect 30 fps interaction effect) thetheoretical value of the bandwidth could be 2980 KBps6036 KBps and 3713 KBps respectively In summary theFPS value one of the most important parameters of the 3Dinteractive experience of the studies of [17ndash19]was correlatedreciprocally with the number of simultaneous tasks on theserver side (server task-pipelines) server processing abilityand network bandwidth and an obvious delay of interactionwas clearly presented By contrast all of the Internet deviceschosen to test our method could provide an FPS value at30ndash60 and the FPS valuewas not correlatedwith server tasksserver processing ability and network bandwidth but with
the device GPU (Graphics Processing Unit) Undoubtedlyalmost all GPUs of Internet devices in the market could meetthe demands of our light Slave model rendering Moreoverthe peak value of the bandwidth under the behavior requestfor final image was less than 50KBps and less than 1 KBpsat idle (including the operation of Slave model) thus bothwere far below those tests of methods of [17ndash19] Thusour method could be more suitable for a complex mobileInternet environment Moreover the complex compressiontransmission methods of [17ndash19] were built on customizedsoftware of plug-ins for the client which is hard to applyto pure web pages and falls short of wide compatibility Bycontrast that is our methodrsquos advantage
For the server load because the Master volume in ourmethod only responds to the final behavior request (instruc-tions) instead of the continuous request of the studies of[9 17ndash19] the test showed that server CPUutilizationwas lessthan 20 at idle (including the operation of Slavemodel) andless than 30 for responding to behavior request thus it wasfar below of those tests ofmethods of [17ndash19] with nearly a fullload during all the operation time under the same conditionsof SMI data quality of final required image server config-uration and volume rendering algorithm Theoretically ourplatform can respond to more independent tasks (pipelines)at the same time Table 2 shows that while the server wasrunning four independent tasks simultaneously the responsetime of our method remained at approximately 1 s in the 100-KBps client bandwidth much better than [17ndash19] which waswith more than 16 secondsrsquo delay of end frame in the singletask status In fact because user viewing and processing ofthe 3D medical volume occurred via interaction on the Slavemodel the delay experience wasmuch fewer than that in [17ndash19] as detailed in Section 41That indicates ourmethod couldhave some more advantages including much less networkload and server load to serve much more independent tasksand users
In addition [17ndash19] did not propose an effective inter-active method to clip the 3D volume what our method hasaccomplished
6 Conclusion
This paper proposed aMaster-Slave dual-channel interactionmode which absorbed the advantages of the surface ren-dering of WebGL and remote volume rendering to builda visualization interaction platform to rapidly view andprocess 3Dmedical volume reconstruction in a pure webpageenvironment This platform could be used in a diversifiednetwork environment (100-KBps bandwidth for extreme)and diversified client devices (no special hardware and oper-ating system constraints) without any special installation butwith a good user experience similar to native applicationThese features could serve authorized users to convenientlyaccess SMIs and present 3D visualization anywhere andcould be a technological basis of online communicationamong doctors and patients [34] Or at least it could helpclinicians to arbitrarily acquire 3D image according to theirclinical needs instead of the static pictures that radiologistsdrafted For example clinicians could conveniently observe
12 International Journal of Telemedicine and Applications
the complicated relationship of skull base bone and bloodvessel via this platform (Figure 6(d)) in the past this pieceof information was provided by a few static images generatedfrom radiologistrsquos understanding
Additionally this solution is designed to be followed asanHTML5 standard interface indicating any authorized usercould rapidly access the application via one URL hyperlinkand the platform could be easily integrated into othertelemedical systems seamlessly as a third-party applicationfor example the Supplementary Materials (2 Demonstrationof the operation procedure on laptop linked by emailmp4and 3 Demonstration of the operation procedure on iPadvia websitemp4) These features may play an assistant roleto improve work patterns and extend the range of applicationof traditional medical image systems For example it can beapplied to regional health telemedicine andmedical imagingeducation at a low cost because there is no special cost ofdeploy and upgrade of software network and hardware onthe client side And more importantly a teacher can vividlyshow the students the image characteristics of the diseaseanywhere a clinician can intuitively introduce disease to thepatients anywhere and ask expertsrsquo help anywhere That maypromote the development of mobile medical technology to acertain degree
Conflicts of Interest
There are no conflicts of interest declared
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China (no 81171866) the National Key BasicResearch Program of China (no 2014CB541602) and theSocial Livelihood Science andTechnology Innovation SpecialProject of CSTC (no cstc2015shmszx120002)
References
[1] Q Zhang M Alexander and L Ryner ldquoSynchronized 2D3Doptical mapping for interactive exploration and real-time visu-alization of multi-function neurological imagesrdquo ComputerizedMedical Imaging and Graphics vol 37 no 7-8 pp 552ndash567 2013
[2] S Zimeras and L G Gortzis ldquoInteractive tele-radiological seg-mentation systems for treatment and diagnosisrdquo InternationalJournal of Telemedicine and Applications vol 2012 Article ID713739 p 15 2012
[3] S E Mahmoudi A Akhondi-Asl R Rahmani et al ldquoWeb-based interactive 2D3D medical image processing andvisualization softwarerdquo Computer Methods and Programs inBiomedicine vol 98 no 2 pp 172ndash182 2010
[4] L Faggioni E Neri I Bargellini et al ldquoIPad-based primary2D reading of CT angiography examinations of patients withsuspected acute gastrointestinal bleeding preliminary experi-ence 1Lrdquo British Journal of Radiology vol 88 no 1047 ArticleID 20140477 2015
[5] J B Park H J Choi J H Lee and B S Kang ldquoAn assessment ofthe iPad 2 as a CT teleradiology tool using brain CT with sub-tle intracranial hemorrhage under conventional illuminationrdquoJournal of Digital Imaging vol 26 no 4 pp 683ndash690 2013
[6] P T Johnson S L Zimmerman D Heath et al ldquoThe iPad asa mobile device for CT display and interpretation diagnosticaccuracy for identification of pulmonary embolismrdquo EmergencyRadiology vol 19 no 4 pp 323ndash327 2012
[7] J M Noguera and J R Jimenez ldquoVisualization of very large 3Dvolumes on mobile devices and WebGLrdquo in Proceedings of the20th International Conference in Central Europe on ComputerGraphics Visualization and Computer Vision WSCG 2012 pp105ndash112 June 2012
[8] J Congote A Segura L Kabongo et al ldquoInteractive visu-alization of volumetric data with WebGL in real-timerdquo inProceedings of the 16th International Conference on 3D WebTechnology (Web3D rsquo11) pp 137ndash145 June 2011
[9] A Schiewe M Anstoots and J Kruger ldquoState of the art inmobile volume rendering on iOS devicesrdquo in Proceedings of theEurographics Conference on Visualization (EuroVis rsquo15) 2015
[10] A Evans M Romeo A Bahrehmand et al ldquo3D graphics on theweb a surveyrdquo Computers amp Graphics vol 41 no 1 pp 43ndash612014
[11] M Callieri R M Andrei M D Benedetto et al ldquoVisualizationmethods formolecular studies on theweb platformrdquo inProceed-ings of the 15th international conference on web 3D technology(Web3D rsquo10) pp 117ndash126 July 2010
[12] J Jimenez A M Lopez J Cruz et al ldquoA Web platform forthe interactive visualization and analysis of the 3D fractaldimension of MRI datardquo Journal of Biomedical Informatics vol51 pp 176ndash190 2014
[13] WebGL httpswwwkhronosorgwebgl[14] C R Butson G Tamm S Jain T Fogal and J Kruger
ldquoEvaluation of interactive visualization on mobile computingplatforms for selection of deep brain stimulation parametersrdquoIEEE Transactions on Visualization and Computer Graphics vol19 no 1 pp 108ndash117 2013
[15] M Levoy ldquoDisplay of surfaces from volume datardquo IEEE Com-puter Graphics and Applications vol 8 no 3 pp 29ndash37 1988
[16] E Kotter T Baumann D Jager and M Langer ldquoTechnologiesfor image distribution in hospitalsrdquo European Radiology vol 16no 6 pp 1270ndash1279 2006
[17] S Park W Kim and I Ihm ldquoMobile collaborative medical dis-play systemrdquo Computer Methods and Programs in Biomedicinevol 89 no 3 pp 248ndash260 2008
[18] J R Mitchell P Sharma J Modi et al ldquoA smartphone client-server teleradiology system for primary diagnosis of acutestrokerdquo Journal of Medical Internet Research vol 13 no 2 p e312011
[19] T Hachaj ldquoReal time exploration and management of largemedical volumetric datasets on small mobile devicesmdasheval-uation of remote volume rendering approachrdquo InternationalJournal of Information Management vol 34 no 3 pp 336ndash3432014
[20] P Quax J Liesenborgs A Barzan et al ldquoRemote renderingsolutions using web technologiesrdquoMultimedia Tools and Appli-cations vol 75 no 8 pp 4383ndash4410 2016
[21] D Preuveneers Y Berbers and W Joosen ldquoThe future ofmobile E-health application development exploring HTML5for context-aware diabetes monitoringrdquo Procedia ComputerScience vol 21 pp 351ndash359 2013
[22] J M Noguera J J Jimenez and M C Osuna-Perez ldquoDevel-opment and evaluation of a 3D mobile application for learningmanual therapy in the physiotherapy laboratoryrdquo Computersand Education vol 69 pp 96ndash108 2013
International Journal of Telemedicine and Applications 13
[23] Y S Kang ldquoVolume rendering overviewrdquo inGPU Programmingand Cg Language Primer chapter 14 pp 154-155 1st edition2009
[24] Visualization Toolkit (VTK) httpwwwvtkorg[25] vtkPiecewiseFunction Class Reference of VTK httpwwwvtk
orgdocnightlyhtmlclassvtkPiecewiseFunctionhtml[26] vtkColorTransferFunction Class Reference of VTK http
wwwvtkorgdocnightlyhtmlannotatedhtml[27] vtkPlane Class Reference of VTK httpwwwvtkorgdoc
nightlyhtmlclassvtkPlanehtml[28] L Ma D-X Zhao and Z-Z Yang ldquoA software tool for
visualization of molecular face (VMF) by improving marchingcubes algorithmrdquo Computational and Theoretical Chemistryvol 1028 pp 34ndash45 2014
[29] G F Wang J Liu and W Liu ldquoDIB display and window trans-formation technology for DICOM medical imagesrdquo ChineseJournal of Medical Device vol 18 no 8 pp 1ndash5 2005
[30] L Qiao X Chen L X Yang et al ldquoDevelopment of medi-cal image three-dimensional-visualization platform based onhome network and pure web conditionsrdquo Beijing BiomedicalEngineering vol 34 no 3 pp 229-233286 2015
[31] M Ying and J Miller ldquoRefactoring legacy AJAX applications toimprove the efficiency of the data exchange componentrdquo Journalof Systems and Software vol 86 no 1 pp 72ndash88 2013
[32] HTML5 httpwwww3orgTR2012CR-html5-20121217[33] QIHU 360 SOFTWARE CO LIMITED 360 Internet Security
httpwww360safecomInternet-securityhtml[34] L Qiao Y Li X Chen et al ldquoMedical high-resolution image
sharing and electronic whiteboard system a pure-web-basedsystem for accessing and discussing lossless original images intelemedicinerdquoComputerMethods and Programs in Biomedicinevol 121 no 2 pp 77ndash91 2015
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
8 International Journal of Telemedicine and Applications
(a) Initial state (b) Viewing angle interaction
(c) Clipping interaction (d) Zooming and coloropacity interaction
Figure 6 Screenshot and operational demonstration
transform the viewing angle from Figure 6(a) to Figure 6(b)(red solid box) and then the submit button is pressed to sendthe behavior request instruction to the server to require anupdated projection image from the Master volume (yellowsolid box) In Figure 6(c) using virtual control panel (reddashed box) to operate Clipping plane after submitting thenew projection image was then updated (left) In Figure 6(d)the mouse wheel or hand gesture is used to zoom the Slavemodel and then the coloropacity theme is submitted via thedropdown list (bottom right corner) to guide remote volumerendering
4 Mode of Availability of the System
In this paper we designed aMaster-Slave dual-channel inter-action mode to improve a problem of remote visualizationinteraction of the volume reconstruction of SMIs whichusually have a huge data size and calculation cost Finallya visualization interaction platform has been built Particu-larly there is no special software and hardware requirementfor the clients and no special network environment isneeded
Tomeasure whether the effectiveness and performance toadapt to the mobile Internet application has characteristicsof lightweight three radiologists with more than 3 years ofexperience were invited to test the interactive experienceimaging quality and compatibility over diversified Internetdevices Moreover four junior students majoring in Biomed-ical Engineeringwere invited to test and quantize the networkload and response time
TestingData 3 groups of SMIs from the radiology departmentof Southwest Hospital of Chongqing include the following
(1) head and neck CT with 495 slices 512 lowast 512 2494MB(2) trunk CT with 609 slices 512 lowast 512 3065MB and (3)thorax andmandible CT with 500 slices 512lowast512 2516MBServer An IBM X3650M4 workstation (CPU 4 times E5-2603RAM 16GB OS Windows Server 2008) was connectedto the Internet via 10Mbps The 2D projection image wasset as the resolution of 512 lowast 512 with high quality JPEGcompression (level-8)
Client 4 Internet devices were located at the 3G or WIFImobile Internet with the server In Figure 7(a) there is aFounder E520 personal computer running on winXP withFirefox 46 web browser via 3G usb card In Figure 7(b) thereis an Acer E15 laptop running on win7 with chrome 43 webbrowser at WIFI In Figure 7(c) there is a HuaWei Honor6 smart phone running on EMUI 30 which is developedfrom Android 442 with the Opera 37 web browser at 3Gnet In Figure 7(d) there is an iPad mini running on IOS6with safari at WIFI All of them could be representative inthe mobile Internet Additionally each was limited to a 100-KBps wireless bandwidth by 360 security firewall for extremetest [33]
41 Interactive Experience Imaging Quality and Compatibil-ity To ensure the consistency of the context the testing data(1) were considered as the main object of reference Here thedemonstration and effect of the testing data (1) are shown inFigure 6
Figure 6 shows that this Slave model of the testing data(1) which was compressed to 104MB could be fit to thedemands of interactive navigation to accurately guide remotevolume rendering and shows that the projection image couldbe displayed to keep the original quality
International Journal of Telemedicine and Applications 9
(a) Personal computer (b) Laptop computer
(c) Smartphone (d) iPad
Figure 7 The system is accessed via four different devices and web browsers under extreme bandwidth
For further validation three radiologists were invited totest the platform with the similar behaviors in Figure 6 Thisjob was repeated to operate three groups of different testingdata through different web browsers and Internet devicesthat were all aforementioned Additionally a five-pointscale was designed to respond to the radiologistsrsquo requests(Table 2)
Table 2 presents a very positive evaluation from the statis-tical result of the imaging quality and interactive experienceand all three radiologists expressed that is an interestingway whether as a gesture-based or mouse-based interactionNo installations were required from the client and goodcompatibility was obtained (same result is shown in Table 3)However one radiologist thought the default Slave modelof the testing data (1) with a skeleton model of 104MB(Figure 3(d2)) was a little rough He suggested to use theskeleton model of 526MB or skin model of 913 KB (Figures3(c1) and 3(c2)) instead of the 104 model both of which arethe alternatives from the mutual interface for the client asdetailed in Section 231 Additionally all three radiologistsclaimed that the response time for final high quality image isalways about 1 second that is consistent with the test resultin Section 42 However they did accept the performancebecause they could directly operate the Slavemodel smoothlywithout any delay on client side thus a short wait for finalhigh quality image had little influence Moreover one of theradiologists specifically expressed the native application of3D rendering on desktop PC also needs 1 or 2 secondsrsquo delayfor final image this performance of the solution running onsmart phone is similar to the native application on PC and isvery good
42 Response Time and Network Load between the Clientand Server The response time can be defined as the timeinterval between the user sending a behavior request tothe system and web browser loading the response resultscompletely The network load can be defined as the amountof data transmission of each interactive behavior Four juniorstudents majoring in Biomedical Engineering were invitedto simultaneously test the platform for one hour During thetest all of the students respectively viewed and processed thetesting data (1) through the platform and performed similarbehaviors to those detailed in Figure 6This test was repeatedby each student with different web browsers and Internetdevices all of which were aforementioned Here we trainedthe students to record the real-time network traffic by the 360security firewall and the time interval by the script log Theresults are showed in Table 3
In Table 3 for each device the amount of data trans-mission and response time of the first loading are muchhigher than the behavior request and idle status because thefirst loading asks the client to download the necessary Slavemodel The response time of the first loading is lt12 s at 100-KBps client bandwidth which is similar to the total timespent on Slavemodel downloading In fact the response timeof first loading also includes the cost time of SMI loadingand rendering on the server side but the server monitorpresents the cost time as lt7 s because SMI loading and Slavemodel downloading are simultaneous and this cost could beneglected
After the first loading the network load of the behaviorrequest for final image is spent on request instructions sub-mission final projection image downloading and the serverprocessing status querying at a fixed frequency of 300ms
10 International Journal of Telemedicine and Applications
Table 2 Statistic results of the five-point scale concerning the interactive experience and imaging quality
Focus Subject Votes from five-point scaleVeryagree Agree Uncertainty Disagree Very
disagree
Interactive experienceMouse-basedGesture-based 998771
The operation of viewing and processing isaccurate
3998771 3
Simple and easy to use without training 2 1998771 3
A good user experience similar to nativeapplication
2 1998771 2 1
Imaging quality
Slavemodel could meet demands of navigation 2 12D image fromMaster volume could provide highquality reconstruction result via utilizing thedisplay capability of Internet terminals
3
Others Latency time for final high quality images 3
Table 3 Average response time and amount of data transmission from the different behaviors and devices at 100-KBps
Devices
First loading (averagevalue)
Behavior request for finalimage (average value)
Idle status ofbandwidthoccupation(KBps)
Frames PerSecond (FPS)Amount of
data trans (KB)Response time(seconds)
Amount ofdata trans (KB)
Response time(seconds)
PC (3G net) 108544 1123 4710 112 023 48ndash60Laptop (WIFI) 106496 1051 4608 089 031 48ndash60iPad mini (WIFI) 107941 1093 4403 095 035 50ndash60Smartphone (3G net) 107502 1072 3994 102 028 30ndash40
where the projection image downloading is the main partThe results present a complete amount of data transmission ofeach behavior request is close to 50KB but the correspondingresponse time for final image is approximately 1 s in 100KBpsThat is because besides the network load the rest of timeis spent on server processing and time interval of querying(300ms) The short delay could be acceptable for practicalapplications as detailed in Section 41
While in the idle status including the userrsquos interactivebehavior (operation) on the Slave model the bandwidthoccupation is almost negligible Moreover there are no sig-nificant differences among diversified devices and networksat 100-KBps client bandwidth
Additionally the FPS of Slave model rendering on theclient side is measured via statsminjs which is a librarypackaged from WebGL When the client rendering is com-pleted FPS could remain at the upper limit
5 Discussion and Comparison
Due to the nonprivate network environment and diversifiedpurposes application software on one device must competefor limited resources of the network and computing Thuslightweight which requires lower resources of the networkand computing and avoids unnecessary installation could bean important characteristic of mobile Internet applicationAdditionally to avoid compatibility problems caused by dif-ferent operating systems and client hardware characteristics
using HTML5 and related web technologies running onmajor web browsers may be a choice to adapt to cross-platform deployment and the characteristics of lightweight[21]
Therefore this article described research on a methodfor the Master-Slave dual-channel interaction mode whichis based on the HTML5 standard (also a wider web com-patibility standard of WebGL) to permit users viewing andprocessing million-megabyte-class SMIs via different clientdevices (without hardware or operating system constraints)over a pure web page (without installing special software) andin a diversified network environment
Compared with the traditional PACS mode of ldquoimagecompression transmission local reconstructionrdquo becauseour method does not transfer original SMIs to the clientit neither needs to wait for a long time to download norrequests special demands of the software environment onthe client device Moreover compared with the state ofthe art of local volume rendering based on WebGL andHTML5 technology [7ndash9] our methods have no specialdemands of client hardware for computation that may reducebattery consumption [9] and enhance the compatibility ofthe client devices our methods may be more fit for mobileInternet application Additionally the client does not accessthe original SMIs which could be much safer
Compared with popular solutions based on absolutesurface rendering technology [11 12 14] our method haslocally used this technique to only support light Slavemodel
International Journal of Telemedicine and Applications 11
for interactive navigation therefore the demands of thenetwork bandwidth and device rendering capability arealmost negligible in ourmethodMoreover the final requiredimage of high quality is a projection drawing from volumerendering technology indicating that our method could beconvenient for users to view the internal anatomy structureand complex organization relationship however surfacerendering technology could not support these characteristicsof volume
Compared with solutions based on absolute remote vol-ume rendering methods our method adopted the Master-Slave dual-channel interaction mode to strengthen the userexperience which is an entirely different approach from themode of ldquocontinuous request continuous responserdquo of studiesof [9] Although studies of [17ndash19] proposed several methodsto improve the mode to reduce the network load includingvideo-compressed transmission [17] tile-based transmission[18] and variable resolution transmission [19] the pressureof server load and real-time data stream increased We usedthe same SMI data (testing data (1)) the same quality offinal required image and the same server configurationand volume rendering algorithm in Section 4 to test themethods in [17ndash19] For quantitative test we formulated aninteractive trajectory with fixed 30 request points theoreti-cally assuming 30 rendering requests triggered per 1 secondthere should be 30 fps render effects On this basis [17]adopted 512 lowast 512 mpeg-4 compression for transmission[18] adopted 8 lowast 8 blocks to segment each 512 lowast 512 frameand transfer the blocks which are different from previousframe in the process of interaction and [19] adopted 64 lowast64 resolution transmission in the process of interaction but512 lowast 512 for end frame The test was repeated for 5 timesunder the condition of one single client monopolizing thebandwidth and server under ideal WIFI environment Theresults showed that the total amount of data transmissionfor each interaction test of [17ndash19] was 2980 KB 6036 KBand 3713 KB respectively the average response time of firstframe was 110 seconds 059 seconds and 052 seconds andthe average cumulative response time of end frame (30thframe which is the final required image) was 2131 seconds1761 seconds and 1601 seconds The response time containsthe factors of server continuous rendering response servercontinuous frame-encoding and so forth (more than 70server CPU utilization during interaction) but the networkfactors could be tentatively neglected under the ideal single-user WIFI environment Assuming that we can improvethe serverrsquos processing ability to completely eliminate theserver response delay (perfect 30 fps interaction effect) thetheoretical value of the bandwidth could be 2980 KBps6036 KBps and 3713 KBps respectively In summary theFPS value one of the most important parameters of the 3Dinteractive experience of the studies of [17ndash19]was correlatedreciprocally with the number of simultaneous tasks on theserver side (server task-pipelines) server processing abilityand network bandwidth and an obvious delay of interactionwas clearly presented By contrast all of the Internet deviceschosen to test our method could provide an FPS value at30ndash60 and the FPS valuewas not correlatedwith server tasksserver processing ability and network bandwidth but with
the device GPU (Graphics Processing Unit) Undoubtedlyalmost all GPUs of Internet devices in the market could meetthe demands of our light Slave model rendering Moreoverthe peak value of the bandwidth under the behavior requestfor final image was less than 50KBps and less than 1 KBpsat idle (including the operation of Slave model) thus bothwere far below those tests of methods of [17ndash19] Thusour method could be more suitable for a complex mobileInternet environment Moreover the complex compressiontransmission methods of [17ndash19] were built on customizedsoftware of plug-ins for the client which is hard to applyto pure web pages and falls short of wide compatibility Bycontrast that is our methodrsquos advantage
For the server load because the Master volume in ourmethod only responds to the final behavior request (instruc-tions) instead of the continuous request of the studies of[9 17ndash19] the test showed that server CPUutilizationwas lessthan 20 at idle (including the operation of Slavemodel) andless than 30 for responding to behavior request thus it wasfar below of those tests ofmethods of [17ndash19] with nearly a fullload during all the operation time under the same conditionsof SMI data quality of final required image server config-uration and volume rendering algorithm Theoretically ourplatform can respond to more independent tasks (pipelines)at the same time Table 2 shows that while the server wasrunning four independent tasks simultaneously the responsetime of our method remained at approximately 1 s in the 100-KBps client bandwidth much better than [17ndash19] which waswith more than 16 secondsrsquo delay of end frame in the singletask status In fact because user viewing and processing ofthe 3D medical volume occurred via interaction on the Slavemodel the delay experience wasmuch fewer than that in [17ndash19] as detailed in Section 41That indicates ourmethod couldhave some more advantages including much less networkload and server load to serve much more independent tasksand users
In addition [17ndash19] did not propose an effective inter-active method to clip the 3D volume what our method hasaccomplished
6 Conclusion
This paper proposed aMaster-Slave dual-channel interactionmode which absorbed the advantages of the surface ren-dering of WebGL and remote volume rendering to builda visualization interaction platform to rapidly view andprocess 3Dmedical volume reconstruction in a pure webpageenvironment This platform could be used in a diversifiednetwork environment (100-KBps bandwidth for extreme)and diversified client devices (no special hardware and oper-ating system constraints) without any special installation butwith a good user experience similar to native applicationThese features could serve authorized users to convenientlyaccess SMIs and present 3D visualization anywhere andcould be a technological basis of online communicationamong doctors and patients [34] Or at least it could helpclinicians to arbitrarily acquire 3D image according to theirclinical needs instead of the static pictures that radiologistsdrafted For example clinicians could conveniently observe
12 International Journal of Telemedicine and Applications
the complicated relationship of skull base bone and bloodvessel via this platform (Figure 6(d)) in the past this pieceof information was provided by a few static images generatedfrom radiologistrsquos understanding
Additionally this solution is designed to be followed asanHTML5 standard interface indicating any authorized usercould rapidly access the application via one URL hyperlinkand the platform could be easily integrated into othertelemedical systems seamlessly as a third-party applicationfor example the Supplementary Materials (2 Demonstrationof the operation procedure on laptop linked by emailmp4and 3 Demonstration of the operation procedure on iPadvia websitemp4) These features may play an assistant roleto improve work patterns and extend the range of applicationof traditional medical image systems For example it can beapplied to regional health telemedicine andmedical imagingeducation at a low cost because there is no special cost ofdeploy and upgrade of software network and hardware onthe client side And more importantly a teacher can vividlyshow the students the image characteristics of the diseaseanywhere a clinician can intuitively introduce disease to thepatients anywhere and ask expertsrsquo help anywhere That maypromote the development of mobile medical technology to acertain degree
Conflicts of Interest
There are no conflicts of interest declared
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China (no 81171866) the National Key BasicResearch Program of China (no 2014CB541602) and theSocial Livelihood Science andTechnology Innovation SpecialProject of CSTC (no cstc2015shmszx120002)
References
[1] Q Zhang M Alexander and L Ryner ldquoSynchronized 2D3Doptical mapping for interactive exploration and real-time visu-alization of multi-function neurological imagesrdquo ComputerizedMedical Imaging and Graphics vol 37 no 7-8 pp 552ndash567 2013
[2] S Zimeras and L G Gortzis ldquoInteractive tele-radiological seg-mentation systems for treatment and diagnosisrdquo InternationalJournal of Telemedicine and Applications vol 2012 Article ID713739 p 15 2012
[3] S E Mahmoudi A Akhondi-Asl R Rahmani et al ldquoWeb-based interactive 2D3D medical image processing andvisualization softwarerdquo Computer Methods and Programs inBiomedicine vol 98 no 2 pp 172ndash182 2010
[4] L Faggioni E Neri I Bargellini et al ldquoIPad-based primary2D reading of CT angiography examinations of patients withsuspected acute gastrointestinal bleeding preliminary experi-ence 1Lrdquo British Journal of Radiology vol 88 no 1047 ArticleID 20140477 2015
[5] J B Park H J Choi J H Lee and B S Kang ldquoAn assessment ofthe iPad 2 as a CT teleradiology tool using brain CT with sub-tle intracranial hemorrhage under conventional illuminationrdquoJournal of Digital Imaging vol 26 no 4 pp 683ndash690 2013
[6] P T Johnson S L Zimmerman D Heath et al ldquoThe iPad asa mobile device for CT display and interpretation diagnosticaccuracy for identification of pulmonary embolismrdquo EmergencyRadiology vol 19 no 4 pp 323ndash327 2012
[7] J M Noguera and J R Jimenez ldquoVisualization of very large 3Dvolumes on mobile devices and WebGLrdquo in Proceedings of the20th International Conference in Central Europe on ComputerGraphics Visualization and Computer Vision WSCG 2012 pp105ndash112 June 2012
[8] J Congote A Segura L Kabongo et al ldquoInteractive visu-alization of volumetric data with WebGL in real-timerdquo inProceedings of the 16th International Conference on 3D WebTechnology (Web3D rsquo11) pp 137ndash145 June 2011
[9] A Schiewe M Anstoots and J Kruger ldquoState of the art inmobile volume rendering on iOS devicesrdquo in Proceedings of theEurographics Conference on Visualization (EuroVis rsquo15) 2015
[10] A Evans M Romeo A Bahrehmand et al ldquo3D graphics on theweb a surveyrdquo Computers amp Graphics vol 41 no 1 pp 43ndash612014
[11] M Callieri R M Andrei M D Benedetto et al ldquoVisualizationmethods formolecular studies on theweb platformrdquo inProceed-ings of the 15th international conference on web 3D technology(Web3D rsquo10) pp 117ndash126 July 2010
[12] J Jimenez A M Lopez J Cruz et al ldquoA Web platform forthe interactive visualization and analysis of the 3D fractaldimension of MRI datardquo Journal of Biomedical Informatics vol51 pp 176ndash190 2014
[13] WebGL httpswwwkhronosorgwebgl[14] C R Butson G Tamm S Jain T Fogal and J Kruger
ldquoEvaluation of interactive visualization on mobile computingplatforms for selection of deep brain stimulation parametersrdquoIEEE Transactions on Visualization and Computer Graphics vol19 no 1 pp 108ndash117 2013
[15] M Levoy ldquoDisplay of surfaces from volume datardquo IEEE Com-puter Graphics and Applications vol 8 no 3 pp 29ndash37 1988
[16] E Kotter T Baumann D Jager and M Langer ldquoTechnologiesfor image distribution in hospitalsrdquo European Radiology vol 16no 6 pp 1270ndash1279 2006
[17] S Park W Kim and I Ihm ldquoMobile collaborative medical dis-play systemrdquo Computer Methods and Programs in Biomedicinevol 89 no 3 pp 248ndash260 2008
[18] J R Mitchell P Sharma J Modi et al ldquoA smartphone client-server teleradiology system for primary diagnosis of acutestrokerdquo Journal of Medical Internet Research vol 13 no 2 p e312011
[19] T Hachaj ldquoReal time exploration and management of largemedical volumetric datasets on small mobile devicesmdasheval-uation of remote volume rendering approachrdquo InternationalJournal of Information Management vol 34 no 3 pp 336ndash3432014
[20] P Quax J Liesenborgs A Barzan et al ldquoRemote renderingsolutions using web technologiesrdquoMultimedia Tools and Appli-cations vol 75 no 8 pp 4383ndash4410 2016
[21] D Preuveneers Y Berbers and W Joosen ldquoThe future ofmobile E-health application development exploring HTML5for context-aware diabetes monitoringrdquo Procedia ComputerScience vol 21 pp 351ndash359 2013
[22] J M Noguera J J Jimenez and M C Osuna-Perez ldquoDevel-opment and evaluation of a 3D mobile application for learningmanual therapy in the physiotherapy laboratoryrdquo Computersand Education vol 69 pp 96ndash108 2013
International Journal of Telemedicine and Applications 13
[23] Y S Kang ldquoVolume rendering overviewrdquo inGPU Programmingand Cg Language Primer chapter 14 pp 154-155 1st edition2009
[24] Visualization Toolkit (VTK) httpwwwvtkorg[25] vtkPiecewiseFunction Class Reference of VTK httpwwwvtk
orgdocnightlyhtmlclassvtkPiecewiseFunctionhtml[26] vtkColorTransferFunction Class Reference of VTK http
wwwvtkorgdocnightlyhtmlannotatedhtml[27] vtkPlane Class Reference of VTK httpwwwvtkorgdoc
nightlyhtmlclassvtkPlanehtml[28] L Ma D-X Zhao and Z-Z Yang ldquoA software tool for
visualization of molecular face (VMF) by improving marchingcubes algorithmrdquo Computational and Theoretical Chemistryvol 1028 pp 34ndash45 2014
[29] G F Wang J Liu and W Liu ldquoDIB display and window trans-formation technology for DICOM medical imagesrdquo ChineseJournal of Medical Device vol 18 no 8 pp 1ndash5 2005
[30] L Qiao X Chen L X Yang et al ldquoDevelopment of medi-cal image three-dimensional-visualization platform based onhome network and pure web conditionsrdquo Beijing BiomedicalEngineering vol 34 no 3 pp 229-233286 2015
[31] M Ying and J Miller ldquoRefactoring legacy AJAX applications toimprove the efficiency of the data exchange componentrdquo Journalof Systems and Software vol 86 no 1 pp 72ndash88 2013
[32] HTML5 httpwwww3orgTR2012CR-html5-20121217[33] QIHU 360 SOFTWARE CO LIMITED 360 Internet Security
httpwww360safecomInternet-securityhtml[34] L Qiao Y Li X Chen et al ldquoMedical high-resolution image
sharing and electronic whiteboard system a pure-web-basedsystem for accessing and discussing lossless original images intelemedicinerdquoComputerMethods and Programs in Biomedicinevol 121 no 2 pp 77ndash91 2015
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of Telemedicine and Applications 9
(a) Personal computer (b) Laptop computer
(c) Smartphone (d) iPad
Figure 7 The system is accessed via four different devices and web browsers under extreme bandwidth
For further validation three radiologists were invited totest the platform with the similar behaviors in Figure 6 Thisjob was repeated to operate three groups of different testingdata through different web browsers and Internet devicesthat were all aforementioned Additionally a five-pointscale was designed to respond to the radiologistsrsquo requests(Table 2)
Table 2 presents a very positive evaluation from the statis-tical result of the imaging quality and interactive experienceand all three radiologists expressed that is an interestingway whether as a gesture-based or mouse-based interactionNo installations were required from the client and goodcompatibility was obtained (same result is shown in Table 3)However one radiologist thought the default Slave modelof the testing data (1) with a skeleton model of 104MB(Figure 3(d2)) was a little rough He suggested to use theskeleton model of 526MB or skin model of 913 KB (Figures3(c1) and 3(c2)) instead of the 104 model both of which arethe alternatives from the mutual interface for the client asdetailed in Section 231 Additionally all three radiologistsclaimed that the response time for final high quality image isalways about 1 second that is consistent with the test resultin Section 42 However they did accept the performancebecause they could directly operate the Slavemodel smoothlywithout any delay on client side thus a short wait for finalhigh quality image had little influence Moreover one of theradiologists specifically expressed the native application of3D rendering on desktop PC also needs 1 or 2 secondsrsquo delayfor final image this performance of the solution running onsmart phone is similar to the native application on PC and isvery good
42 Response Time and Network Load between the Clientand Server The response time can be defined as the timeinterval between the user sending a behavior request tothe system and web browser loading the response resultscompletely The network load can be defined as the amountof data transmission of each interactive behavior Four juniorstudents majoring in Biomedical Engineering were invitedto simultaneously test the platform for one hour During thetest all of the students respectively viewed and processed thetesting data (1) through the platform and performed similarbehaviors to those detailed in Figure 6This test was repeatedby each student with different web browsers and Internetdevices all of which were aforementioned Here we trainedthe students to record the real-time network traffic by the 360security firewall and the time interval by the script log Theresults are showed in Table 3
In Table 3 for each device the amount of data trans-mission and response time of the first loading are muchhigher than the behavior request and idle status because thefirst loading asks the client to download the necessary Slavemodel The response time of the first loading is lt12 s at 100-KBps client bandwidth which is similar to the total timespent on Slavemodel downloading In fact the response timeof first loading also includes the cost time of SMI loadingand rendering on the server side but the server monitorpresents the cost time as lt7 s because SMI loading and Slavemodel downloading are simultaneous and this cost could beneglected
After the first loading the network load of the behaviorrequest for final image is spent on request instructions sub-mission final projection image downloading and the serverprocessing status querying at a fixed frequency of 300ms
10 International Journal of Telemedicine and Applications
Table 2 Statistic results of the five-point scale concerning the interactive experience and imaging quality
Focus Subject Votes from five-point scaleVeryagree Agree Uncertainty Disagree Very
disagree
Interactive experienceMouse-basedGesture-based 998771
The operation of viewing and processing isaccurate
3998771 3
Simple and easy to use without training 2 1998771 3
A good user experience similar to nativeapplication
2 1998771 2 1
Imaging quality
Slavemodel could meet demands of navigation 2 12D image fromMaster volume could provide highquality reconstruction result via utilizing thedisplay capability of Internet terminals
3
Others Latency time for final high quality images 3
Table 3 Average response time and amount of data transmission from the different behaviors and devices at 100-KBps
Devices
First loading (averagevalue)
Behavior request for finalimage (average value)
Idle status ofbandwidthoccupation(KBps)
Frames PerSecond (FPS)Amount of
data trans (KB)Response time(seconds)
Amount ofdata trans (KB)
Response time(seconds)
PC (3G net) 108544 1123 4710 112 023 48ndash60Laptop (WIFI) 106496 1051 4608 089 031 48ndash60iPad mini (WIFI) 107941 1093 4403 095 035 50ndash60Smartphone (3G net) 107502 1072 3994 102 028 30ndash40
where the projection image downloading is the main partThe results present a complete amount of data transmission ofeach behavior request is close to 50KB but the correspondingresponse time for final image is approximately 1 s in 100KBpsThat is because besides the network load the rest of timeis spent on server processing and time interval of querying(300ms) The short delay could be acceptable for practicalapplications as detailed in Section 41
While in the idle status including the userrsquos interactivebehavior (operation) on the Slave model the bandwidthoccupation is almost negligible Moreover there are no sig-nificant differences among diversified devices and networksat 100-KBps client bandwidth
Additionally the FPS of Slave model rendering on theclient side is measured via statsminjs which is a librarypackaged from WebGL When the client rendering is com-pleted FPS could remain at the upper limit
5 Discussion and Comparison
Due to the nonprivate network environment and diversifiedpurposes application software on one device must competefor limited resources of the network and computing Thuslightweight which requires lower resources of the networkand computing and avoids unnecessary installation could bean important characteristic of mobile Internet applicationAdditionally to avoid compatibility problems caused by dif-ferent operating systems and client hardware characteristics
using HTML5 and related web technologies running onmajor web browsers may be a choice to adapt to cross-platform deployment and the characteristics of lightweight[21]
Therefore this article described research on a methodfor the Master-Slave dual-channel interaction mode whichis based on the HTML5 standard (also a wider web com-patibility standard of WebGL) to permit users viewing andprocessing million-megabyte-class SMIs via different clientdevices (without hardware or operating system constraints)over a pure web page (without installing special software) andin a diversified network environment
Compared with the traditional PACS mode of ldquoimagecompression transmission local reconstructionrdquo becauseour method does not transfer original SMIs to the clientit neither needs to wait for a long time to download norrequests special demands of the software environment onthe client device Moreover compared with the state ofthe art of local volume rendering based on WebGL andHTML5 technology [7ndash9] our methods have no specialdemands of client hardware for computation that may reducebattery consumption [9] and enhance the compatibility ofthe client devices our methods may be more fit for mobileInternet application Additionally the client does not accessthe original SMIs which could be much safer
Compared with popular solutions based on absolutesurface rendering technology [11 12 14] our method haslocally used this technique to only support light Slavemodel
International Journal of Telemedicine and Applications 11
for interactive navigation therefore the demands of thenetwork bandwidth and device rendering capability arealmost negligible in ourmethodMoreover the final requiredimage of high quality is a projection drawing from volumerendering technology indicating that our method could beconvenient for users to view the internal anatomy structureand complex organization relationship however surfacerendering technology could not support these characteristicsof volume
Compared with solutions based on absolute remote vol-ume rendering methods our method adopted the Master-Slave dual-channel interaction mode to strengthen the userexperience which is an entirely different approach from themode of ldquocontinuous request continuous responserdquo of studiesof [9] Although studies of [17ndash19] proposed several methodsto improve the mode to reduce the network load includingvideo-compressed transmission [17] tile-based transmission[18] and variable resolution transmission [19] the pressureof server load and real-time data stream increased We usedthe same SMI data (testing data (1)) the same quality offinal required image and the same server configurationand volume rendering algorithm in Section 4 to test themethods in [17ndash19] For quantitative test we formulated aninteractive trajectory with fixed 30 request points theoreti-cally assuming 30 rendering requests triggered per 1 secondthere should be 30 fps render effects On this basis [17]adopted 512 lowast 512 mpeg-4 compression for transmission[18] adopted 8 lowast 8 blocks to segment each 512 lowast 512 frameand transfer the blocks which are different from previousframe in the process of interaction and [19] adopted 64 lowast64 resolution transmission in the process of interaction but512 lowast 512 for end frame The test was repeated for 5 timesunder the condition of one single client monopolizing thebandwidth and server under ideal WIFI environment Theresults showed that the total amount of data transmissionfor each interaction test of [17ndash19] was 2980 KB 6036 KBand 3713 KB respectively the average response time of firstframe was 110 seconds 059 seconds and 052 seconds andthe average cumulative response time of end frame (30thframe which is the final required image) was 2131 seconds1761 seconds and 1601 seconds The response time containsthe factors of server continuous rendering response servercontinuous frame-encoding and so forth (more than 70server CPU utilization during interaction) but the networkfactors could be tentatively neglected under the ideal single-user WIFI environment Assuming that we can improvethe serverrsquos processing ability to completely eliminate theserver response delay (perfect 30 fps interaction effect) thetheoretical value of the bandwidth could be 2980 KBps6036 KBps and 3713 KBps respectively In summary theFPS value one of the most important parameters of the 3Dinteractive experience of the studies of [17ndash19]was correlatedreciprocally with the number of simultaneous tasks on theserver side (server task-pipelines) server processing abilityand network bandwidth and an obvious delay of interactionwas clearly presented By contrast all of the Internet deviceschosen to test our method could provide an FPS value at30ndash60 and the FPS valuewas not correlatedwith server tasksserver processing ability and network bandwidth but with
the device GPU (Graphics Processing Unit) Undoubtedlyalmost all GPUs of Internet devices in the market could meetthe demands of our light Slave model rendering Moreoverthe peak value of the bandwidth under the behavior requestfor final image was less than 50KBps and less than 1 KBpsat idle (including the operation of Slave model) thus bothwere far below those tests of methods of [17ndash19] Thusour method could be more suitable for a complex mobileInternet environment Moreover the complex compressiontransmission methods of [17ndash19] were built on customizedsoftware of plug-ins for the client which is hard to applyto pure web pages and falls short of wide compatibility Bycontrast that is our methodrsquos advantage
For the server load because the Master volume in ourmethod only responds to the final behavior request (instruc-tions) instead of the continuous request of the studies of[9 17ndash19] the test showed that server CPUutilizationwas lessthan 20 at idle (including the operation of Slavemodel) andless than 30 for responding to behavior request thus it wasfar below of those tests ofmethods of [17ndash19] with nearly a fullload during all the operation time under the same conditionsof SMI data quality of final required image server config-uration and volume rendering algorithm Theoretically ourplatform can respond to more independent tasks (pipelines)at the same time Table 2 shows that while the server wasrunning four independent tasks simultaneously the responsetime of our method remained at approximately 1 s in the 100-KBps client bandwidth much better than [17ndash19] which waswith more than 16 secondsrsquo delay of end frame in the singletask status In fact because user viewing and processing ofthe 3D medical volume occurred via interaction on the Slavemodel the delay experience wasmuch fewer than that in [17ndash19] as detailed in Section 41That indicates ourmethod couldhave some more advantages including much less networkload and server load to serve much more independent tasksand users
In addition [17ndash19] did not propose an effective inter-active method to clip the 3D volume what our method hasaccomplished
6 Conclusion
This paper proposed aMaster-Slave dual-channel interactionmode which absorbed the advantages of the surface ren-dering of WebGL and remote volume rendering to builda visualization interaction platform to rapidly view andprocess 3Dmedical volume reconstruction in a pure webpageenvironment This platform could be used in a diversifiednetwork environment (100-KBps bandwidth for extreme)and diversified client devices (no special hardware and oper-ating system constraints) without any special installation butwith a good user experience similar to native applicationThese features could serve authorized users to convenientlyaccess SMIs and present 3D visualization anywhere andcould be a technological basis of online communicationamong doctors and patients [34] Or at least it could helpclinicians to arbitrarily acquire 3D image according to theirclinical needs instead of the static pictures that radiologistsdrafted For example clinicians could conveniently observe
12 International Journal of Telemedicine and Applications
the complicated relationship of skull base bone and bloodvessel via this platform (Figure 6(d)) in the past this pieceof information was provided by a few static images generatedfrom radiologistrsquos understanding
Additionally this solution is designed to be followed asanHTML5 standard interface indicating any authorized usercould rapidly access the application via one URL hyperlinkand the platform could be easily integrated into othertelemedical systems seamlessly as a third-party applicationfor example the Supplementary Materials (2 Demonstrationof the operation procedure on laptop linked by emailmp4and 3 Demonstration of the operation procedure on iPadvia websitemp4) These features may play an assistant roleto improve work patterns and extend the range of applicationof traditional medical image systems For example it can beapplied to regional health telemedicine andmedical imagingeducation at a low cost because there is no special cost ofdeploy and upgrade of software network and hardware onthe client side And more importantly a teacher can vividlyshow the students the image characteristics of the diseaseanywhere a clinician can intuitively introduce disease to thepatients anywhere and ask expertsrsquo help anywhere That maypromote the development of mobile medical technology to acertain degree
Conflicts of Interest
There are no conflicts of interest declared
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China (no 81171866) the National Key BasicResearch Program of China (no 2014CB541602) and theSocial Livelihood Science andTechnology Innovation SpecialProject of CSTC (no cstc2015shmszx120002)
References
[1] Q Zhang M Alexander and L Ryner ldquoSynchronized 2D3Doptical mapping for interactive exploration and real-time visu-alization of multi-function neurological imagesrdquo ComputerizedMedical Imaging and Graphics vol 37 no 7-8 pp 552ndash567 2013
[2] S Zimeras and L G Gortzis ldquoInteractive tele-radiological seg-mentation systems for treatment and diagnosisrdquo InternationalJournal of Telemedicine and Applications vol 2012 Article ID713739 p 15 2012
[3] S E Mahmoudi A Akhondi-Asl R Rahmani et al ldquoWeb-based interactive 2D3D medical image processing andvisualization softwarerdquo Computer Methods and Programs inBiomedicine vol 98 no 2 pp 172ndash182 2010
[4] L Faggioni E Neri I Bargellini et al ldquoIPad-based primary2D reading of CT angiography examinations of patients withsuspected acute gastrointestinal bleeding preliminary experi-ence 1Lrdquo British Journal of Radiology vol 88 no 1047 ArticleID 20140477 2015
[5] J B Park H J Choi J H Lee and B S Kang ldquoAn assessment ofthe iPad 2 as a CT teleradiology tool using brain CT with sub-tle intracranial hemorrhage under conventional illuminationrdquoJournal of Digital Imaging vol 26 no 4 pp 683ndash690 2013
[6] P T Johnson S L Zimmerman D Heath et al ldquoThe iPad asa mobile device for CT display and interpretation diagnosticaccuracy for identification of pulmonary embolismrdquo EmergencyRadiology vol 19 no 4 pp 323ndash327 2012
[7] J M Noguera and J R Jimenez ldquoVisualization of very large 3Dvolumes on mobile devices and WebGLrdquo in Proceedings of the20th International Conference in Central Europe on ComputerGraphics Visualization and Computer Vision WSCG 2012 pp105ndash112 June 2012
[8] J Congote A Segura L Kabongo et al ldquoInteractive visu-alization of volumetric data with WebGL in real-timerdquo inProceedings of the 16th International Conference on 3D WebTechnology (Web3D rsquo11) pp 137ndash145 June 2011
[9] A Schiewe M Anstoots and J Kruger ldquoState of the art inmobile volume rendering on iOS devicesrdquo in Proceedings of theEurographics Conference on Visualization (EuroVis rsquo15) 2015
[10] A Evans M Romeo A Bahrehmand et al ldquo3D graphics on theweb a surveyrdquo Computers amp Graphics vol 41 no 1 pp 43ndash612014
[11] M Callieri R M Andrei M D Benedetto et al ldquoVisualizationmethods formolecular studies on theweb platformrdquo inProceed-ings of the 15th international conference on web 3D technology(Web3D rsquo10) pp 117ndash126 July 2010
[12] J Jimenez A M Lopez J Cruz et al ldquoA Web platform forthe interactive visualization and analysis of the 3D fractaldimension of MRI datardquo Journal of Biomedical Informatics vol51 pp 176ndash190 2014
[13] WebGL httpswwwkhronosorgwebgl[14] C R Butson G Tamm S Jain T Fogal and J Kruger
ldquoEvaluation of interactive visualization on mobile computingplatforms for selection of deep brain stimulation parametersrdquoIEEE Transactions on Visualization and Computer Graphics vol19 no 1 pp 108ndash117 2013
[15] M Levoy ldquoDisplay of surfaces from volume datardquo IEEE Com-puter Graphics and Applications vol 8 no 3 pp 29ndash37 1988
[16] E Kotter T Baumann D Jager and M Langer ldquoTechnologiesfor image distribution in hospitalsrdquo European Radiology vol 16no 6 pp 1270ndash1279 2006
[17] S Park W Kim and I Ihm ldquoMobile collaborative medical dis-play systemrdquo Computer Methods and Programs in Biomedicinevol 89 no 3 pp 248ndash260 2008
[18] J R Mitchell P Sharma J Modi et al ldquoA smartphone client-server teleradiology system for primary diagnosis of acutestrokerdquo Journal of Medical Internet Research vol 13 no 2 p e312011
[19] T Hachaj ldquoReal time exploration and management of largemedical volumetric datasets on small mobile devicesmdasheval-uation of remote volume rendering approachrdquo InternationalJournal of Information Management vol 34 no 3 pp 336ndash3432014
[20] P Quax J Liesenborgs A Barzan et al ldquoRemote renderingsolutions using web technologiesrdquoMultimedia Tools and Appli-cations vol 75 no 8 pp 4383ndash4410 2016
[21] D Preuveneers Y Berbers and W Joosen ldquoThe future ofmobile E-health application development exploring HTML5for context-aware diabetes monitoringrdquo Procedia ComputerScience vol 21 pp 351ndash359 2013
[22] J M Noguera J J Jimenez and M C Osuna-Perez ldquoDevel-opment and evaluation of a 3D mobile application for learningmanual therapy in the physiotherapy laboratoryrdquo Computersand Education vol 69 pp 96ndash108 2013
International Journal of Telemedicine and Applications 13
[23] Y S Kang ldquoVolume rendering overviewrdquo inGPU Programmingand Cg Language Primer chapter 14 pp 154-155 1st edition2009
[24] Visualization Toolkit (VTK) httpwwwvtkorg[25] vtkPiecewiseFunction Class Reference of VTK httpwwwvtk
orgdocnightlyhtmlclassvtkPiecewiseFunctionhtml[26] vtkColorTransferFunction Class Reference of VTK http
wwwvtkorgdocnightlyhtmlannotatedhtml[27] vtkPlane Class Reference of VTK httpwwwvtkorgdoc
nightlyhtmlclassvtkPlanehtml[28] L Ma D-X Zhao and Z-Z Yang ldquoA software tool for
visualization of molecular face (VMF) by improving marchingcubes algorithmrdquo Computational and Theoretical Chemistryvol 1028 pp 34ndash45 2014
[29] G F Wang J Liu and W Liu ldquoDIB display and window trans-formation technology for DICOM medical imagesrdquo ChineseJournal of Medical Device vol 18 no 8 pp 1ndash5 2005
[30] L Qiao X Chen L X Yang et al ldquoDevelopment of medi-cal image three-dimensional-visualization platform based onhome network and pure web conditionsrdquo Beijing BiomedicalEngineering vol 34 no 3 pp 229-233286 2015
[31] M Ying and J Miller ldquoRefactoring legacy AJAX applications toimprove the efficiency of the data exchange componentrdquo Journalof Systems and Software vol 86 no 1 pp 72ndash88 2013
[32] HTML5 httpwwww3orgTR2012CR-html5-20121217[33] QIHU 360 SOFTWARE CO LIMITED 360 Internet Security
httpwww360safecomInternet-securityhtml[34] L Qiao Y Li X Chen et al ldquoMedical high-resolution image
sharing and electronic whiteboard system a pure-web-basedsystem for accessing and discussing lossless original images intelemedicinerdquoComputerMethods and Programs in Biomedicinevol 121 no 2 pp 77ndash91 2015
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
10 International Journal of Telemedicine and Applications
Table 2 Statistic results of the five-point scale concerning the interactive experience and imaging quality
Focus Subject Votes from five-point scaleVeryagree Agree Uncertainty Disagree Very
disagree
Interactive experienceMouse-basedGesture-based 998771
The operation of viewing and processing isaccurate
3998771 3
Simple and easy to use without training 2 1998771 3
A good user experience similar to nativeapplication
2 1998771 2 1
Imaging quality
Slavemodel could meet demands of navigation 2 12D image fromMaster volume could provide highquality reconstruction result via utilizing thedisplay capability of Internet terminals
3
Others Latency time for final high quality images 3
Table 3 Average response time and amount of data transmission from the different behaviors and devices at 100-KBps
Devices
First loading (averagevalue)
Behavior request for finalimage (average value)
Idle status ofbandwidthoccupation(KBps)
Frames PerSecond (FPS)Amount of
data trans (KB)Response time(seconds)
Amount ofdata trans (KB)
Response time(seconds)
PC (3G net) 108544 1123 4710 112 023 48ndash60Laptop (WIFI) 106496 1051 4608 089 031 48ndash60iPad mini (WIFI) 107941 1093 4403 095 035 50ndash60Smartphone (3G net) 107502 1072 3994 102 028 30ndash40
where the projection image downloading is the main partThe results present a complete amount of data transmission ofeach behavior request is close to 50KB but the correspondingresponse time for final image is approximately 1 s in 100KBpsThat is because besides the network load the rest of timeis spent on server processing and time interval of querying(300ms) The short delay could be acceptable for practicalapplications as detailed in Section 41
While in the idle status including the userrsquos interactivebehavior (operation) on the Slave model the bandwidthoccupation is almost negligible Moreover there are no sig-nificant differences among diversified devices and networksat 100-KBps client bandwidth
Additionally the FPS of Slave model rendering on theclient side is measured via statsminjs which is a librarypackaged from WebGL When the client rendering is com-pleted FPS could remain at the upper limit
5 Discussion and Comparison
Due to the nonprivate network environment and diversifiedpurposes application software on one device must competefor limited resources of the network and computing Thuslightweight which requires lower resources of the networkand computing and avoids unnecessary installation could bean important characteristic of mobile Internet applicationAdditionally to avoid compatibility problems caused by dif-ferent operating systems and client hardware characteristics
using HTML5 and related web technologies running onmajor web browsers may be a choice to adapt to cross-platform deployment and the characteristics of lightweight[21]
Therefore this article described research on a methodfor the Master-Slave dual-channel interaction mode whichis based on the HTML5 standard (also a wider web com-patibility standard of WebGL) to permit users viewing andprocessing million-megabyte-class SMIs via different clientdevices (without hardware or operating system constraints)over a pure web page (without installing special software) andin a diversified network environment
Compared with the traditional PACS mode of ldquoimagecompression transmission local reconstructionrdquo becauseour method does not transfer original SMIs to the clientit neither needs to wait for a long time to download norrequests special demands of the software environment onthe client device Moreover compared with the state ofthe art of local volume rendering based on WebGL andHTML5 technology [7ndash9] our methods have no specialdemands of client hardware for computation that may reducebattery consumption [9] and enhance the compatibility ofthe client devices our methods may be more fit for mobileInternet application Additionally the client does not accessthe original SMIs which could be much safer
Compared with popular solutions based on absolutesurface rendering technology [11 12 14] our method haslocally used this technique to only support light Slavemodel
International Journal of Telemedicine and Applications 11
for interactive navigation therefore the demands of thenetwork bandwidth and device rendering capability arealmost negligible in ourmethodMoreover the final requiredimage of high quality is a projection drawing from volumerendering technology indicating that our method could beconvenient for users to view the internal anatomy structureand complex organization relationship however surfacerendering technology could not support these characteristicsof volume
Compared with solutions based on absolute remote vol-ume rendering methods our method adopted the Master-Slave dual-channel interaction mode to strengthen the userexperience which is an entirely different approach from themode of ldquocontinuous request continuous responserdquo of studiesof [9] Although studies of [17ndash19] proposed several methodsto improve the mode to reduce the network load includingvideo-compressed transmission [17] tile-based transmission[18] and variable resolution transmission [19] the pressureof server load and real-time data stream increased We usedthe same SMI data (testing data (1)) the same quality offinal required image and the same server configurationand volume rendering algorithm in Section 4 to test themethods in [17ndash19] For quantitative test we formulated aninteractive trajectory with fixed 30 request points theoreti-cally assuming 30 rendering requests triggered per 1 secondthere should be 30 fps render effects On this basis [17]adopted 512 lowast 512 mpeg-4 compression for transmission[18] adopted 8 lowast 8 blocks to segment each 512 lowast 512 frameand transfer the blocks which are different from previousframe in the process of interaction and [19] adopted 64 lowast64 resolution transmission in the process of interaction but512 lowast 512 for end frame The test was repeated for 5 timesunder the condition of one single client monopolizing thebandwidth and server under ideal WIFI environment Theresults showed that the total amount of data transmissionfor each interaction test of [17ndash19] was 2980 KB 6036 KBand 3713 KB respectively the average response time of firstframe was 110 seconds 059 seconds and 052 seconds andthe average cumulative response time of end frame (30thframe which is the final required image) was 2131 seconds1761 seconds and 1601 seconds The response time containsthe factors of server continuous rendering response servercontinuous frame-encoding and so forth (more than 70server CPU utilization during interaction) but the networkfactors could be tentatively neglected under the ideal single-user WIFI environment Assuming that we can improvethe serverrsquos processing ability to completely eliminate theserver response delay (perfect 30 fps interaction effect) thetheoretical value of the bandwidth could be 2980 KBps6036 KBps and 3713 KBps respectively In summary theFPS value one of the most important parameters of the 3Dinteractive experience of the studies of [17ndash19]was correlatedreciprocally with the number of simultaneous tasks on theserver side (server task-pipelines) server processing abilityand network bandwidth and an obvious delay of interactionwas clearly presented By contrast all of the Internet deviceschosen to test our method could provide an FPS value at30ndash60 and the FPS valuewas not correlatedwith server tasksserver processing ability and network bandwidth but with
the device GPU (Graphics Processing Unit) Undoubtedlyalmost all GPUs of Internet devices in the market could meetthe demands of our light Slave model rendering Moreoverthe peak value of the bandwidth under the behavior requestfor final image was less than 50KBps and less than 1 KBpsat idle (including the operation of Slave model) thus bothwere far below those tests of methods of [17ndash19] Thusour method could be more suitable for a complex mobileInternet environment Moreover the complex compressiontransmission methods of [17ndash19] were built on customizedsoftware of plug-ins for the client which is hard to applyto pure web pages and falls short of wide compatibility Bycontrast that is our methodrsquos advantage
For the server load because the Master volume in ourmethod only responds to the final behavior request (instruc-tions) instead of the continuous request of the studies of[9 17ndash19] the test showed that server CPUutilizationwas lessthan 20 at idle (including the operation of Slavemodel) andless than 30 for responding to behavior request thus it wasfar below of those tests ofmethods of [17ndash19] with nearly a fullload during all the operation time under the same conditionsof SMI data quality of final required image server config-uration and volume rendering algorithm Theoretically ourplatform can respond to more independent tasks (pipelines)at the same time Table 2 shows that while the server wasrunning four independent tasks simultaneously the responsetime of our method remained at approximately 1 s in the 100-KBps client bandwidth much better than [17ndash19] which waswith more than 16 secondsrsquo delay of end frame in the singletask status In fact because user viewing and processing ofthe 3D medical volume occurred via interaction on the Slavemodel the delay experience wasmuch fewer than that in [17ndash19] as detailed in Section 41That indicates ourmethod couldhave some more advantages including much less networkload and server load to serve much more independent tasksand users
In addition [17ndash19] did not propose an effective inter-active method to clip the 3D volume what our method hasaccomplished
6 Conclusion
This paper proposed aMaster-Slave dual-channel interactionmode which absorbed the advantages of the surface ren-dering of WebGL and remote volume rendering to builda visualization interaction platform to rapidly view andprocess 3Dmedical volume reconstruction in a pure webpageenvironment This platform could be used in a diversifiednetwork environment (100-KBps bandwidth for extreme)and diversified client devices (no special hardware and oper-ating system constraints) without any special installation butwith a good user experience similar to native applicationThese features could serve authorized users to convenientlyaccess SMIs and present 3D visualization anywhere andcould be a technological basis of online communicationamong doctors and patients [34] Or at least it could helpclinicians to arbitrarily acquire 3D image according to theirclinical needs instead of the static pictures that radiologistsdrafted For example clinicians could conveniently observe
12 International Journal of Telemedicine and Applications
the complicated relationship of skull base bone and bloodvessel via this platform (Figure 6(d)) in the past this pieceof information was provided by a few static images generatedfrom radiologistrsquos understanding
Additionally this solution is designed to be followed asanHTML5 standard interface indicating any authorized usercould rapidly access the application via one URL hyperlinkand the platform could be easily integrated into othertelemedical systems seamlessly as a third-party applicationfor example the Supplementary Materials (2 Demonstrationof the operation procedure on laptop linked by emailmp4and 3 Demonstration of the operation procedure on iPadvia websitemp4) These features may play an assistant roleto improve work patterns and extend the range of applicationof traditional medical image systems For example it can beapplied to regional health telemedicine andmedical imagingeducation at a low cost because there is no special cost ofdeploy and upgrade of software network and hardware onthe client side And more importantly a teacher can vividlyshow the students the image characteristics of the diseaseanywhere a clinician can intuitively introduce disease to thepatients anywhere and ask expertsrsquo help anywhere That maypromote the development of mobile medical technology to acertain degree
Conflicts of Interest
There are no conflicts of interest declared
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China (no 81171866) the National Key BasicResearch Program of China (no 2014CB541602) and theSocial Livelihood Science andTechnology Innovation SpecialProject of CSTC (no cstc2015shmszx120002)
References
[1] Q Zhang M Alexander and L Ryner ldquoSynchronized 2D3Doptical mapping for interactive exploration and real-time visu-alization of multi-function neurological imagesrdquo ComputerizedMedical Imaging and Graphics vol 37 no 7-8 pp 552ndash567 2013
[2] S Zimeras and L G Gortzis ldquoInteractive tele-radiological seg-mentation systems for treatment and diagnosisrdquo InternationalJournal of Telemedicine and Applications vol 2012 Article ID713739 p 15 2012
[3] S E Mahmoudi A Akhondi-Asl R Rahmani et al ldquoWeb-based interactive 2D3D medical image processing andvisualization softwarerdquo Computer Methods and Programs inBiomedicine vol 98 no 2 pp 172ndash182 2010
[4] L Faggioni E Neri I Bargellini et al ldquoIPad-based primary2D reading of CT angiography examinations of patients withsuspected acute gastrointestinal bleeding preliminary experi-ence 1Lrdquo British Journal of Radiology vol 88 no 1047 ArticleID 20140477 2015
[5] J B Park H J Choi J H Lee and B S Kang ldquoAn assessment ofthe iPad 2 as a CT teleradiology tool using brain CT with sub-tle intracranial hemorrhage under conventional illuminationrdquoJournal of Digital Imaging vol 26 no 4 pp 683ndash690 2013
[6] P T Johnson S L Zimmerman D Heath et al ldquoThe iPad asa mobile device for CT display and interpretation diagnosticaccuracy for identification of pulmonary embolismrdquo EmergencyRadiology vol 19 no 4 pp 323ndash327 2012
[7] J M Noguera and J R Jimenez ldquoVisualization of very large 3Dvolumes on mobile devices and WebGLrdquo in Proceedings of the20th International Conference in Central Europe on ComputerGraphics Visualization and Computer Vision WSCG 2012 pp105ndash112 June 2012
[8] J Congote A Segura L Kabongo et al ldquoInteractive visu-alization of volumetric data with WebGL in real-timerdquo inProceedings of the 16th International Conference on 3D WebTechnology (Web3D rsquo11) pp 137ndash145 June 2011
[9] A Schiewe M Anstoots and J Kruger ldquoState of the art inmobile volume rendering on iOS devicesrdquo in Proceedings of theEurographics Conference on Visualization (EuroVis rsquo15) 2015
[10] A Evans M Romeo A Bahrehmand et al ldquo3D graphics on theweb a surveyrdquo Computers amp Graphics vol 41 no 1 pp 43ndash612014
[11] M Callieri R M Andrei M D Benedetto et al ldquoVisualizationmethods formolecular studies on theweb platformrdquo inProceed-ings of the 15th international conference on web 3D technology(Web3D rsquo10) pp 117ndash126 July 2010
[12] J Jimenez A M Lopez J Cruz et al ldquoA Web platform forthe interactive visualization and analysis of the 3D fractaldimension of MRI datardquo Journal of Biomedical Informatics vol51 pp 176ndash190 2014
[13] WebGL httpswwwkhronosorgwebgl[14] C R Butson G Tamm S Jain T Fogal and J Kruger
ldquoEvaluation of interactive visualization on mobile computingplatforms for selection of deep brain stimulation parametersrdquoIEEE Transactions on Visualization and Computer Graphics vol19 no 1 pp 108ndash117 2013
[15] M Levoy ldquoDisplay of surfaces from volume datardquo IEEE Com-puter Graphics and Applications vol 8 no 3 pp 29ndash37 1988
[16] E Kotter T Baumann D Jager and M Langer ldquoTechnologiesfor image distribution in hospitalsrdquo European Radiology vol 16no 6 pp 1270ndash1279 2006
[17] S Park W Kim and I Ihm ldquoMobile collaborative medical dis-play systemrdquo Computer Methods and Programs in Biomedicinevol 89 no 3 pp 248ndash260 2008
[18] J R Mitchell P Sharma J Modi et al ldquoA smartphone client-server teleradiology system for primary diagnosis of acutestrokerdquo Journal of Medical Internet Research vol 13 no 2 p e312011
[19] T Hachaj ldquoReal time exploration and management of largemedical volumetric datasets on small mobile devicesmdasheval-uation of remote volume rendering approachrdquo InternationalJournal of Information Management vol 34 no 3 pp 336ndash3432014
[20] P Quax J Liesenborgs A Barzan et al ldquoRemote renderingsolutions using web technologiesrdquoMultimedia Tools and Appli-cations vol 75 no 8 pp 4383ndash4410 2016
[21] D Preuveneers Y Berbers and W Joosen ldquoThe future ofmobile E-health application development exploring HTML5for context-aware diabetes monitoringrdquo Procedia ComputerScience vol 21 pp 351ndash359 2013
[22] J M Noguera J J Jimenez and M C Osuna-Perez ldquoDevel-opment and evaluation of a 3D mobile application for learningmanual therapy in the physiotherapy laboratoryrdquo Computersand Education vol 69 pp 96ndash108 2013
International Journal of Telemedicine and Applications 13
[23] Y S Kang ldquoVolume rendering overviewrdquo inGPU Programmingand Cg Language Primer chapter 14 pp 154-155 1st edition2009
[24] Visualization Toolkit (VTK) httpwwwvtkorg[25] vtkPiecewiseFunction Class Reference of VTK httpwwwvtk
orgdocnightlyhtmlclassvtkPiecewiseFunctionhtml[26] vtkColorTransferFunction Class Reference of VTK http
wwwvtkorgdocnightlyhtmlannotatedhtml[27] vtkPlane Class Reference of VTK httpwwwvtkorgdoc
nightlyhtmlclassvtkPlanehtml[28] L Ma D-X Zhao and Z-Z Yang ldquoA software tool for
visualization of molecular face (VMF) by improving marchingcubes algorithmrdquo Computational and Theoretical Chemistryvol 1028 pp 34ndash45 2014
[29] G F Wang J Liu and W Liu ldquoDIB display and window trans-formation technology for DICOM medical imagesrdquo ChineseJournal of Medical Device vol 18 no 8 pp 1ndash5 2005
[30] L Qiao X Chen L X Yang et al ldquoDevelopment of medi-cal image three-dimensional-visualization platform based onhome network and pure web conditionsrdquo Beijing BiomedicalEngineering vol 34 no 3 pp 229-233286 2015
[31] M Ying and J Miller ldquoRefactoring legacy AJAX applications toimprove the efficiency of the data exchange componentrdquo Journalof Systems and Software vol 86 no 1 pp 72ndash88 2013
[32] HTML5 httpwwww3orgTR2012CR-html5-20121217[33] QIHU 360 SOFTWARE CO LIMITED 360 Internet Security
httpwww360safecomInternet-securityhtml[34] L Qiao Y Li X Chen et al ldquoMedical high-resolution image
sharing and electronic whiteboard system a pure-web-basedsystem for accessing and discussing lossless original images intelemedicinerdquoComputerMethods and Programs in Biomedicinevol 121 no 2 pp 77ndash91 2015
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of Telemedicine and Applications 11
for interactive navigation therefore the demands of thenetwork bandwidth and device rendering capability arealmost negligible in ourmethodMoreover the final requiredimage of high quality is a projection drawing from volumerendering technology indicating that our method could beconvenient for users to view the internal anatomy structureand complex organization relationship however surfacerendering technology could not support these characteristicsof volume
Compared with solutions based on absolute remote vol-ume rendering methods our method adopted the Master-Slave dual-channel interaction mode to strengthen the userexperience which is an entirely different approach from themode of ldquocontinuous request continuous responserdquo of studiesof [9] Although studies of [17ndash19] proposed several methodsto improve the mode to reduce the network load includingvideo-compressed transmission [17] tile-based transmission[18] and variable resolution transmission [19] the pressureof server load and real-time data stream increased We usedthe same SMI data (testing data (1)) the same quality offinal required image and the same server configurationand volume rendering algorithm in Section 4 to test themethods in [17ndash19] For quantitative test we formulated aninteractive trajectory with fixed 30 request points theoreti-cally assuming 30 rendering requests triggered per 1 secondthere should be 30 fps render effects On this basis [17]adopted 512 lowast 512 mpeg-4 compression for transmission[18] adopted 8 lowast 8 blocks to segment each 512 lowast 512 frameand transfer the blocks which are different from previousframe in the process of interaction and [19] adopted 64 lowast64 resolution transmission in the process of interaction but512 lowast 512 for end frame The test was repeated for 5 timesunder the condition of one single client monopolizing thebandwidth and server under ideal WIFI environment Theresults showed that the total amount of data transmissionfor each interaction test of [17ndash19] was 2980 KB 6036 KBand 3713 KB respectively the average response time of firstframe was 110 seconds 059 seconds and 052 seconds andthe average cumulative response time of end frame (30thframe which is the final required image) was 2131 seconds1761 seconds and 1601 seconds The response time containsthe factors of server continuous rendering response servercontinuous frame-encoding and so forth (more than 70server CPU utilization during interaction) but the networkfactors could be tentatively neglected under the ideal single-user WIFI environment Assuming that we can improvethe serverrsquos processing ability to completely eliminate theserver response delay (perfect 30 fps interaction effect) thetheoretical value of the bandwidth could be 2980 KBps6036 KBps and 3713 KBps respectively In summary theFPS value one of the most important parameters of the 3Dinteractive experience of the studies of [17ndash19]was correlatedreciprocally with the number of simultaneous tasks on theserver side (server task-pipelines) server processing abilityand network bandwidth and an obvious delay of interactionwas clearly presented By contrast all of the Internet deviceschosen to test our method could provide an FPS value at30ndash60 and the FPS valuewas not correlatedwith server tasksserver processing ability and network bandwidth but with
the device GPU (Graphics Processing Unit) Undoubtedlyalmost all GPUs of Internet devices in the market could meetthe demands of our light Slave model rendering Moreoverthe peak value of the bandwidth under the behavior requestfor final image was less than 50KBps and less than 1 KBpsat idle (including the operation of Slave model) thus bothwere far below those tests of methods of [17ndash19] Thusour method could be more suitable for a complex mobileInternet environment Moreover the complex compressiontransmission methods of [17ndash19] were built on customizedsoftware of plug-ins for the client which is hard to applyto pure web pages and falls short of wide compatibility Bycontrast that is our methodrsquos advantage
For the server load because the Master volume in ourmethod only responds to the final behavior request (instruc-tions) instead of the continuous request of the studies of[9 17ndash19] the test showed that server CPUutilizationwas lessthan 20 at idle (including the operation of Slavemodel) andless than 30 for responding to behavior request thus it wasfar below of those tests ofmethods of [17ndash19] with nearly a fullload during all the operation time under the same conditionsof SMI data quality of final required image server config-uration and volume rendering algorithm Theoretically ourplatform can respond to more independent tasks (pipelines)at the same time Table 2 shows that while the server wasrunning four independent tasks simultaneously the responsetime of our method remained at approximately 1 s in the 100-KBps client bandwidth much better than [17ndash19] which waswith more than 16 secondsrsquo delay of end frame in the singletask status In fact because user viewing and processing ofthe 3D medical volume occurred via interaction on the Slavemodel the delay experience wasmuch fewer than that in [17ndash19] as detailed in Section 41That indicates ourmethod couldhave some more advantages including much less networkload and server load to serve much more independent tasksand users
In addition [17ndash19] did not propose an effective inter-active method to clip the 3D volume what our method hasaccomplished
6 Conclusion
This paper proposed aMaster-Slave dual-channel interactionmode which absorbed the advantages of the surface ren-dering of WebGL and remote volume rendering to builda visualization interaction platform to rapidly view andprocess 3Dmedical volume reconstruction in a pure webpageenvironment This platform could be used in a diversifiednetwork environment (100-KBps bandwidth for extreme)and diversified client devices (no special hardware and oper-ating system constraints) without any special installation butwith a good user experience similar to native applicationThese features could serve authorized users to convenientlyaccess SMIs and present 3D visualization anywhere andcould be a technological basis of online communicationamong doctors and patients [34] Or at least it could helpclinicians to arbitrarily acquire 3D image according to theirclinical needs instead of the static pictures that radiologistsdrafted For example clinicians could conveniently observe
12 International Journal of Telemedicine and Applications
the complicated relationship of skull base bone and bloodvessel via this platform (Figure 6(d)) in the past this pieceof information was provided by a few static images generatedfrom radiologistrsquos understanding
Additionally this solution is designed to be followed asanHTML5 standard interface indicating any authorized usercould rapidly access the application via one URL hyperlinkand the platform could be easily integrated into othertelemedical systems seamlessly as a third-party applicationfor example the Supplementary Materials (2 Demonstrationof the operation procedure on laptop linked by emailmp4and 3 Demonstration of the operation procedure on iPadvia websitemp4) These features may play an assistant roleto improve work patterns and extend the range of applicationof traditional medical image systems For example it can beapplied to regional health telemedicine andmedical imagingeducation at a low cost because there is no special cost ofdeploy and upgrade of software network and hardware onthe client side And more importantly a teacher can vividlyshow the students the image characteristics of the diseaseanywhere a clinician can intuitively introduce disease to thepatients anywhere and ask expertsrsquo help anywhere That maypromote the development of mobile medical technology to acertain degree
Conflicts of Interest
There are no conflicts of interest declared
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China (no 81171866) the National Key BasicResearch Program of China (no 2014CB541602) and theSocial Livelihood Science andTechnology Innovation SpecialProject of CSTC (no cstc2015shmszx120002)
References
[1] Q Zhang M Alexander and L Ryner ldquoSynchronized 2D3Doptical mapping for interactive exploration and real-time visu-alization of multi-function neurological imagesrdquo ComputerizedMedical Imaging and Graphics vol 37 no 7-8 pp 552ndash567 2013
[2] S Zimeras and L G Gortzis ldquoInteractive tele-radiological seg-mentation systems for treatment and diagnosisrdquo InternationalJournal of Telemedicine and Applications vol 2012 Article ID713739 p 15 2012
[3] S E Mahmoudi A Akhondi-Asl R Rahmani et al ldquoWeb-based interactive 2D3D medical image processing andvisualization softwarerdquo Computer Methods and Programs inBiomedicine vol 98 no 2 pp 172ndash182 2010
[4] L Faggioni E Neri I Bargellini et al ldquoIPad-based primary2D reading of CT angiography examinations of patients withsuspected acute gastrointestinal bleeding preliminary experi-ence 1Lrdquo British Journal of Radiology vol 88 no 1047 ArticleID 20140477 2015
[5] J B Park H J Choi J H Lee and B S Kang ldquoAn assessment ofthe iPad 2 as a CT teleradiology tool using brain CT with sub-tle intracranial hemorrhage under conventional illuminationrdquoJournal of Digital Imaging vol 26 no 4 pp 683ndash690 2013
[6] P T Johnson S L Zimmerman D Heath et al ldquoThe iPad asa mobile device for CT display and interpretation diagnosticaccuracy for identification of pulmonary embolismrdquo EmergencyRadiology vol 19 no 4 pp 323ndash327 2012
[7] J M Noguera and J R Jimenez ldquoVisualization of very large 3Dvolumes on mobile devices and WebGLrdquo in Proceedings of the20th International Conference in Central Europe on ComputerGraphics Visualization and Computer Vision WSCG 2012 pp105ndash112 June 2012
[8] J Congote A Segura L Kabongo et al ldquoInteractive visu-alization of volumetric data with WebGL in real-timerdquo inProceedings of the 16th International Conference on 3D WebTechnology (Web3D rsquo11) pp 137ndash145 June 2011
[9] A Schiewe M Anstoots and J Kruger ldquoState of the art inmobile volume rendering on iOS devicesrdquo in Proceedings of theEurographics Conference on Visualization (EuroVis rsquo15) 2015
[10] A Evans M Romeo A Bahrehmand et al ldquo3D graphics on theweb a surveyrdquo Computers amp Graphics vol 41 no 1 pp 43ndash612014
[11] M Callieri R M Andrei M D Benedetto et al ldquoVisualizationmethods formolecular studies on theweb platformrdquo inProceed-ings of the 15th international conference on web 3D technology(Web3D rsquo10) pp 117ndash126 July 2010
[12] J Jimenez A M Lopez J Cruz et al ldquoA Web platform forthe interactive visualization and analysis of the 3D fractaldimension of MRI datardquo Journal of Biomedical Informatics vol51 pp 176ndash190 2014
[13] WebGL httpswwwkhronosorgwebgl[14] C R Butson G Tamm S Jain T Fogal and J Kruger
ldquoEvaluation of interactive visualization on mobile computingplatforms for selection of deep brain stimulation parametersrdquoIEEE Transactions on Visualization and Computer Graphics vol19 no 1 pp 108ndash117 2013
[15] M Levoy ldquoDisplay of surfaces from volume datardquo IEEE Com-puter Graphics and Applications vol 8 no 3 pp 29ndash37 1988
[16] E Kotter T Baumann D Jager and M Langer ldquoTechnologiesfor image distribution in hospitalsrdquo European Radiology vol 16no 6 pp 1270ndash1279 2006
[17] S Park W Kim and I Ihm ldquoMobile collaborative medical dis-play systemrdquo Computer Methods and Programs in Biomedicinevol 89 no 3 pp 248ndash260 2008
[18] J R Mitchell P Sharma J Modi et al ldquoA smartphone client-server teleradiology system for primary diagnosis of acutestrokerdquo Journal of Medical Internet Research vol 13 no 2 p e312011
[19] T Hachaj ldquoReal time exploration and management of largemedical volumetric datasets on small mobile devicesmdasheval-uation of remote volume rendering approachrdquo InternationalJournal of Information Management vol 34 no 3 pp 336ndash3432014
[20] P Quax J Liesenborgs A Barzan et al ldquoRemote renderingsolutions using web technologiesrdquoMultimedia Tools and Appli-cations vol 75 no 8 pp 4383ndash4410 2016
[21] D Preuveneers Y Berbers and W Joosen ldquoThe future ofmobile E-health application development exploring HTML5for context-aware diabetes monitoringrdquo Procedia ComputerScience vol 21 pp 351ndash359 2013
[22] J M Noguera J J Jimenez and M C Osuna-Perez ldquoDevel-opment and evaluation of a 3D mobile application for learningmanual therapy in the physiotherapy laboratoryrdquo Computersand Education vol 69 pp 96ndash108 2013
International Journal of Telemedicine and Applications 13
[23] Y S Kang ldquoVolume rendering overviewrdquo inGPU Programmingand Cg Language Primer chapter 14 pp 154-155 1st edition2009
[24] Visualization Toolkit (VTK) httpwwwvtkorg[25] vtkPiecewiseFunction Class Reference of VTK httpwwwvtk
orgdocnightlyhtmlclassvtkPiecewiseFunctionhtml[26] vtkColorTransferFunction Class Reference of VTK http
wwwvtkorgdocnightlyhtmlannotatedhtml[27] vtkPlane Class Reference of VTK httpwwwvtkorgdoc
nightlyhtmlclassvtkPlanehtml[28] L Ma D-X Zhao and Z-Z Yang ldquoA software tool for
visualization of molecular face (VMF) by improving marchingcubes algorithmrdquo Computational and Theoretical Chemistryvol 1028 pp 34ndash45 2014
[29] G F Wang J Liu and W Liu ldquoDIB display and window trans-formation technology for DICOM medical imagesrdquo ChineseJournal of Medical Device vol 18 no 8 pp 1ndash5 2005
[30] L Qiao X Chen L X Yang et al ldquoDevelopment of medi-cal image three-dimensional-visualization platform based onhome network and pure web conditionsrdquo Beijing BiomedicalEngineering vol 34 no 3 pp 229-233286 2015
[31] M Ying and J Miller ldquoRefactoring legacy AJAX applications toimprove the efficiency of the data exchange componentrdquo Journalof Systems and Software vol 86 no 1 pp 72ndash88 2013
[32] HTML5 httpwwww3orgTR2012CR-html5-20121217[33] QIHU 360 SOFTWARE CO LIMITED 360 Internet Security
httpwww360safecomInternet-securityhtml[34] L Qiao Y Li X Chen et al ldquoMedical high-resolution image
sharing and electronic whiteboard system a pure-web-basedsystem for accessing and discussing lossless original images intelemedicinerdquoComputerMethods and Programs in Biomedicinevol 121 no 2 pp 77ndash91 2015
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
12 International Journal of Telemedicine and Applications
the complicated relationship of skull base bone and bloodvessel via this platform (Figure 6(d)) in the past this pieceof information was provided by a few static images generatedfrom radiologistrsquos understanding
Additionally this solution is designed to be followed asanHTML5 standard interface indicating any authorized usercould rapidly access the application via one URL hyperlinkand the platform could be easily integrated into othertelemedical systems seamlessly as a third-party applicationfor example the Supplementary Materials (2 Demonstrationof the operation procedure on laptop linked by emailmp4and 3 Demonstration of the operation procedure on iPadvia websitemp4) These features may play an assistant roleto improve work patterns and extend the range of applicationof traditional medical image systems For example it can beapplied to regional health telemedicine andmedical imagingeducation at a low cost because there is no special cost ofdeploy and upgrade of software network and hardware onthe client side And more importantly a teacher can vividlyshow the students the image characteristics of the diseaseanywhere a clinician can intuitively introduce disease to thepatients anywhere and ask expertsrsquo help anywhere That maypromote the development of mobile medical technology to acertain degree
Conflicts of Interest
There are no conflicts of interest declared
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China (no 81171866) the National Key BasicResearch Program of China (no 2014CB541602) and theSocial Livelihood Science andTechnology Innovation SpecialProject of CSTC (no cstc2015shmszx120002)
References
[1] Q Zhang M Alexander and L Ryner ldquoSynchronized 2D3Doptical mapping for interactive exploration and real-time visu-alization of multi-function neurological imagesrdquo ComputerizedMedical Imaging and Graphics vol 37 no 7-8 pp 552ndash567 2013
[2] S Zimeras and L G Gortzis ldquoInteractive tele-radiological seg-mentation systems for treatment and diagnosisrdquo InternationalJournal of Telemedicine and Applications vol 2012 Article ID713739 p 15 2012
[3] S E Mahmoudi A Akhondi-Asl R Rahmani et al ldquoWeb-based interactive 2D3D medical image processing andvisualization softwarerdquo Computer Methods and Programs inBiomedicine vol 98 no 2 pp 172ndash182 2010
[4] L Faggioni E Neri I Bargellini et al ldquoIPad-based primary2D reading of CT angiography examinations of patients withsuspected acute gastrointestinal bleeding preliminary experi-ence 1Lrdquo British Journal of Radiology vol 88 no 1047 ArticleID 20140477 2015
[5] J B Park H J Choi J H Lee and B S Kang ldquoAn assessment ofthe iPad 2 as a CT teleradiology tool using brain CT with sub-tle intracranial hemorrhage under conventional illuminationrdquoJournal of Digital Imaging vol 26 no 4 pp 683ndash690 2013
[6] P T Johnson S L Zimmerman D Heath et al ldquoThe iPad asa mobile device for CT display and interpretation diagnosticaccuracy for identification of pulmonary embolismrdquo EmergencyRadiology vol 19 no 4 pp 323ndash327 2012
[7] J M Noguera and J R Jimenez ldquoVisualization of very large 3Dvolumes on mobile devices and WebGLrdquo in Proceedings of the20th International Conference in Central Europe on ComputerGraphics Visualization and Computer Vision WSCG 2012 pp105ndash112 June 2012
[8] J Congote A Segura L Kabongo et al ldquoInteractive visu-alization of volumetric data with WebGL in real-timerdquo inProceedings of the 16th International Conference on 3D WebTechnology (Web3D rsquo11) pp 137ndash145 June 2011
[9] A Schiewe M Anstoots and J Kruger ldquoState of the art inmobile volume rendering on iOS devicesrdquo in Proceedings of theEurographics Conference on Visualization (EuroVis rsquo15) 2015
[10] A Evans M Romeo A Bahrehmand et al ldquo3D graphics on theweb a surveyrdquo Computers amp Graphics vol 41 no 1 pp 43ndash612014
[11] M Callieri R M Andrei M D Benedetto et al ldquoVisualizationmethods formolecular studies on theweb platformrdquo inProceed-ings of the 15th international conference on web 3D technology(Web3D rsquo10) pp 117ndash126 July 2010
[12] J Jimenez A M Lopez J Cruz et al ldquoA Web platform forthe interactive visualization and analysis of the 3D fractaldimension of MRI datardquo Journal of Biomedical Informatics vol51 pp 176ndash190 2014
[13] WebGL httpswwwkhronosorgwebgl[14] C R Butson G Tamm S Jain T Fogal and J Kruger
ldquoEvaluation of interactive visualization on mobile computingplatforms for selection of deep brain stimulation parametersrdquoIEEE Transactions on Visualization and Computer Graphics vol19 no 1 pp 108ndash117 2013
[15] M Levoy ldquoDisplay of surfaces from volume datardquo IEEE Com-puter Graphics and Applications vol 8 no 3 pp 29ndash37 1988
[16] E Kotter T Baumann D Jager and M Langer ldquoTechnologiesfor image distribution in hospitalsrdquo European Radiology vol 16no 6 pp 1270ndash1279 2006
[17] S Park W Kim and I Ihm ldquoMobile collaborative medical dis-play systemrdquo Computer Methods and Programs in Biomedicinevol 89 no 3 pp 248ndash260 2008
[18] J R Mitchell P Sharma J Modi et al ldquoA smartphone client-server teleradiology system for primary diagnosis of acutestrokerdquo Journal of Medical Internet Research vol 13 no 2 p e312011
[19] T Hachaj ldquoReal time exploration and management of largemedical volumetric datasets on small mobile devicesmdasheval-uation of remote volume rendering approachrdquo InternationalJournal of Information Management vol 34 no 3 pp 336ndash3432014
[20] P Quax J Liesenborgs A Barzan et al ldquoRemote renderingsolutions using web technologiesrdquoMultimedia Tools and Appli-cations vol 75 no 8 pp 4383ndash4410 2016
[21] D Preuveneers Y Berbers and W Joosen ldquoThe future ofmobile E-health application development exploring HTML5for context-aware diabetes monitoringrdquo Procedia ComputerScience vol 21 pp 351ndash359 2013
[22] J M Noguera J J Jimenez and M C Osuna-Perez ldquoDevel-opment and evaluation of a 3D mobile application for learningmanual therapy in the physiotherapy laboratoryrdquo Computersand Education vol 69 pp 96ndash108 2013
International Journal of Telemedicine and Applications 13
[23] Y S Kang ldquoVolume rendering overviewrdquo inGPU Programmingand Cg Language Primer chapter 14 pp 154-155 1st edition2009
[24] Visualization Toolkit (VTK) httpwwwvtkorg[25] vtkPiecewiseFunction Class Reference of VTK httpwwwvtk
orgdocnightlyhtmlclassvtkPiecewiseFunctionhtml[26] vtkColorTransferFunction Class Reference of VTK http
wwwvtkorgdocnightlyhtmlannotatedhtml[27] vtkPlane Class Reference of VTK httpwwwvtkorgdoc
nightlyhtmlclassvtkPlanehtml[28] L Ma D-X Zhao and Z-Z Yang ldquoA software tool for
visualization of molecular face (VMF) by improving marchingcubes algorithmrdquo Computational and Theoretical Chemistryvol 1028 pp 34ndash45 2014
[29] G F Wang J Liu and W Liu ldquoDIB display and window trans-formation technology for DICOM medical imagesrdquo ChineseJournal of Medical Device vol 18 no 8 pp 1ndash5 2005
[30] L Qiao X Chen L X Yang et al ldquoDevelopment of medi-cal image three-dimensional-visualization platform based onhome network and pure web conditionsrdquo Beijing BiomedicalEngineering vol 34 no 3 pp 229-233286 2015
[31] M Ying and J Miller ldquoRefactoring legacy AJAX applications toimprove the efficiency of the data exchange componentrdquo Journalof Systems and Software vol 86 no 1 pp 72ndash88 2013
[32] HTML5 httpwwww3orgTR2012CR-html5-20121217[33] QIHU 360 SOFTWARE CO LIMITED 360 Internet Security
httpwww360safecomInternet-securityhtml[34] L Qiao Y Li X Chen et al ldquoMedical high-resolution image
sharing and electronic whiteboard system a pure-web-basedsystem for accessing and discussing lossless original images intelemedicinerdquoComputerMethods and Programs in Biomedicinevol 121 no 2 pp 77ndash91 2015
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of Telemedicine and Applications 13
[23] Y S Kang ldquoVolume rendering overviewrdquo inGPU Programmingand Cg Language Primer chapter 14 pp 154-155 1st edition2009
[24] Visualization Toolkit (VTK) httpwwwvtkorg[25] vtkPiecewiseFunction Class Reference of VTK httpwwwvtk
orgdocnightlyhtmlclassvtkPiecewiseFunctionhtml[26] vtkColorTransferFunction Class Reference of VTK http
wwwvtkorgdocnightlyhtmlannotatedhtml[27] vtkPlane Class Reference of VTK httpwwwvtkorgdoc
nightlyhtmlclassvtkPlanehtml[28] L Ma D-X Zhao and Z-Z Yang ldquoA software tool for
visualization of molecular face (VMF) by improving marchingcubes algorithmrdquo Computational and Theoretical Chemistryvol 1028 pp 34ndash45 2014
[29] G F Wang J Liu and W Liu ldquoDIB display and window trans-formation technology for DICOM medical imagesrdquo ChineseJournal of Medical Device vol 18 no 8 pp 1ndash5 2005
[30] L Qiao X Chen L X Yang et al ldquoDevelopment of medi-cal image three-dimensional-visualization platform based onhome network and pure web conditionsrdquo Beijing BiomedicalEngineering vol 34 no 3 pp 229-233286 2015
[31] M Ying and J Miller ldquoRefactoring legacy AJAX applications toimprove the efficiency of the data exchange componentrdquo Journalof Systems and Software vol 86 no 1 pp 72ndash88 2013
[32] HTML5 httpwwww3orgTR2012CR-html5-20121217[33] QIHU 360 SOFTWARE CO LIMITED 360 Internet Security
httpwww360safecomInternet-securityhtml[34] L Qiao Y Li X Chen et al ldquoMedical high-resolution image
sharing and electronic whiteboard system a pure-web-basedsystem for accessing and discussing lossless original images intelemedicinerdquoComputerMethods and Programs in Biomedicinevol 121 no 2 pp 77ndash91 2015
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of