Estimating Motion From MRI Dataee-classes.usc.edu/ee691/read/Ozturk2003p1268.pdfthe user a known pattern to detect for motion tracking. The MRI signal is a modulated complex quantity,

Estimating Motion From MRI Data

CENGIZHAN OZTURK, MEMBER, IEEE, J. ANDREW DERBYSHIRE,AND

ELLIOT R. MCVEIGH, MEMBER, IEEE

Invited Paper

Magnetic resonance imaging (MRI) is an ideal imaging modalityto measure blood flow and tissue motion. It provides excellent con-trast between soft tissues, and images can be acquired at posi-tions and orientations freely defined by the user. From a temporalsequence of MR images, boundaries and edges of tissues can betracked by image processing techniques. Additionally, MRI permitsthe source of the image signal to be manipulated. For example, tem-porary magnetic tags displaying a pattern of variable brightnessmay be placed in the object using MR saturation techniques, givingthe user a known pattern to detect for motion tracking. The MRIsignal is a modulated complex quantity, being derived from a ro-tating magnetic field in the form of an induced current. Well-definedpatterns can also be introduced into the phase of the magnetization,and could be thought of as generalized tags. If the phase of eachpixel is preserved during image reconstruction, relative phase shiftscan be used to directly encode displacement, velocity and acceler-ation. New methods for modeling motion fields from MRI have nowfound application in cardiovascular and other soft tissue imaging.In this review, we shall describe the methods used for encoding,imaging, and modeling motion fields with MRI.

Keywords—Cardiac magnetic resonance, displacement en-coding with stimulated echoes (DENSE), harmonic phase imaging(HARP), magnetic resonance imaging (MRI), magnetic resonancetagging, motion tracking, phase contrast magnetic resonanceimaging, review.

I. INTRODUCTION

There are two main driving forces behind the researchin motion analysis using magnetic resonance imaging(MRI): 1) the quantitative measurement of blood flow,; and

Manuscript received July 15, 2002; revised March 17, 2003.C. Ozturk is with the Institute of Biomedical Engineering, Bogazici

University, Istanbul, Turkey, and also with the National Institutes ofHealth, National Heart, Lung, and Blood Institute (NHLBI), Bethesda, MD20892-1538 USA (e-mail: [email protected]).

J. A. Derbyshire is with the National Institutes of Health, National Heart,Lung, and Blood Institute (NHLBI), Bethesda, MD 20892-1061 USA(e-mail: [email protected]).

E. R. McVeigh is with the National Institutes of Health, NationalHeart, Lung, and Blood Institute, Bethesda, MD 20892-1061 USA andalso with the Department of Biomedical Engineering, Johns HopkinsUniversity School of Medicine, Baltimore MD 21205 USA (e-mail:[email protected]).

Digital Object Identifier 10.1109/JPROC.2003.817872

2) the analysis of heart and other tissue motion (tongue,lung, upper airway, bones, and connective tissues at a jointand skeletal muscle). For direct application to medicine,the measurement of myocardial function is tremendouslyimportant; cardiovascular diseases are the primary causeof mortality in developed countries [1]. Development ofbetter functional cardiac imaging techniques in the hopeof achieving early diagnosis and better patient follow-upis therefore a very active research area. Our focus in thispaper is the assessment of tissue motion using MRI; specialemphasis will be cardiac applications. Due to space limita-tions, we had to leave out two topics: MR elastography [2]imaging and MR diffusion imaging [3].

Although this paper focuses on the heart, there are severalother organ systems where such a detailed motion analysiscan be very helpful and has been pursued. Examples are: 1)dynamic imaging of the musculoskeletal system; 2) investi-gation of other soft tissue (tongue, chest wall, and lung) mo-tion; and 3) detailed motion analysis of internal organs duringcertain exercises.

In the next section we will give a brief background on MRIand several motion analysis techniques. These are: taggedMRI (TMRI), phase contrast MRI (PCMRI), displacementencoding with stimulated echoes (DENSE), and harmonicphase imaging (HARP). Examples are given for each tech-nique in the third section. Comparative analysis, advantagesand disadvantages of each technique, hardware issues, tech-nical challenges, and a brief look into the future can be foundin the final section.

II. BACKGROUND AND THEORETICAL DEVELOPMENT

In this section we start with a brief introduction to MRI.The aim is to introduce only the basics of MRI that arenecessary to understand the main principles of the differentimaging techniques that will be presented in the followingsubsections. For introductory material on MRI, the reader isreferred to [4] and [5]; more theoretical material on cardiacmotion analysis with a detailed survey of current physical

0018-9219/03$17.00 © 2003 IEEE

PROCEEDINGS OF THE IEEE, VOL. 91, NO. 10, OCTOBER 2003 1627

and mathematical models and postprocessing techniques forcardiac MRI can be found in [6].

A. An Introduction to Magnetic Resonance Imaging

MRI [7] is based on exciting magnetic dipoles, mainly pro-tons of water molecules within the body, and observing thesubsequent effects. In MRI, the subject is first introduced intoa high magnetic field with a commonly used field strengthof 1.5 T for clinical imaging (compare it to earth’s mag-netic field strength of 40–70T). Ensembles of protons alignthemselves very quickly with the main field. In the semiclas-sical vector model, the ensembles can be regarded as magne-tization vectors that behave like spinning tops rotating with afrequency proportional to the strength of the local magneticfield , known as the Larmor frequency. The constant

is known as the gyromagnetic ratio and is approximately42.58 MHz/T for protons.

The magnetization can be disturbed (excited, tipped, orflipped) from its aligned equilibrium state by the brief ap-plication of a second magnetic field oriented perpendic-ularly to the main field and tuned to the Larmor frequency.Since this frequency is approximately 64 MHz for protonsat 1.5 T, the applied pulse is often referred to as a radio fre-quency (RF) pulse and can be generated using a tuned coildriven by an RF amplifier. The effect of the RF pulse isto tip the magnetization vectors out of alignment with themain field, providing a transverse component

orthogonal to the main field and a longitudinalcomponent parallel to it, where is theinitial equilibrium longitudinal magnetization and is theflip angle given by . The excited magnetiza-tion precesses, rotating around the main field at the Larmorfrequency like a spinning top rotates around the vertical grav-itational field. The rotating transverse component of the mag-netization induces electromotive force (EMF) in a con-ducting RFreceiver coilplaced near the subject (this couldbe the same coil as used for the excitation pulse), whichis measured as the MR signal by the MRI scanner.

After excitation, the magnetization vectors gradually re-turn to their equilibrium positions, aligned with the mainmagnetic field by processes which can be characterizedusing simple first-order kinetics of the two orthogonal com-ponents of magnetization. The longitudinal magnetizationcomponent recovers exponentially with a rate constant of

, known as the longitudinal or spin-lattice relaxation time,toward the equilibrium state magnetization according to

(1)

The transverse magnetization component decays expo-nentially with a rate constant of , known as the transverseor spin-spin relaxation time, according to

(2)

Various types of tissues (e.g., muscle, fat, blood) often havedifferent relaxation times, providing a source of contrast inMR images. Generally, , but in practice, the MR

signal appears to decay with an even shorter rate constant of. This latter decay time constant arises primarily from the

dephasing effects of magnetic field inhomogeneities (for ex-ample, due to susceptibility differences between tissues) overthe region of interest. These cause the various magnetizationvectors to dephase with respect to one another, resulting insignal attenuation due to destructive interference. Techniquesexist for generating the so-called spin echoes and stimulatedechoes [8] that can reverse much of the effects of this inho-mogeneous decay, in which case the more fundamental decayrate is revealed. The and relaxation parameters areeffective measures of the ability of the magnetization to actas a memory and retain information. Patterns (such as tagsfor motion tracking) encoded at in the longitudinal

and transverse magnetization are lost withrate constants and , respectively.

Magnetic field gradients are of great importance for boththe formation of MR images and motion encoding. Gradientsprovide the means for performing spatially specific opera-tions. The gradient field is a small magnetic field with linearspatial variation that can be superposed on the main fieldunder scanner control. Writing the main field as ,the gradient magnetic field is given by

(3)

Physically, the gradient field is generated by three sets of coilwindings providing linear field variation in each of the phys-ical , , and directions. Gradient fields in any arbitrarydirection are obtained by applying combinations of thethree primary gradients in the ratio of the direction cosinesof . In the presence of the gradient field, the magnetizationprecessing according to the Larmor equation will have a fre-quency offset directly proportional to the component of itsposition in the direction of the magnetic field variation givenby

(4)

Thus, frequency selective operations can be made spatiallyselective through the use of gradients. It is most commonto work in a rotating frame of reference so that frequencyoffsets are measured relative to the Larmor frequency. Thus,the influence of a gradient can be written as

(5)

Slice selection is achieved through the simultaneous appli-cation of an RF pulse and gradient field [9]. The RF pulse ismodulated so that it is band limited and, consequently, ex-cites the magnetization within a specific, limited range offrequencies. The gradient field is applied such that it varieslinearly in the desired slice selection direction: thus, only themagnetization vectors within a planar slab are excited. Theslice profile is determined by the duration and modulation ofthe RF pulse, while its position in the slice gradient directioncan be adjusted by modifying the RF pulse frequency band.The slab thickness and orientation are controlled by adjustingthe amplitude and direction of the slice selection gradient.

1628 PROCEEDINGS OF THE IEEE, VOL. 91, NO. 10, OCTOBER 2003

After excitation, the imposition of further gradient fieldswill cause the magnetization in different regions of the sliceto precess with a spectrum of frequencies determined by thespatial positions of the magnetization vectors in accordancewith (4). This spectrum can be found by acquiring the MRsignal and performing a Fourier transformation on it, therebyproviding a one-dimensional (1-D) spatial profile of the sub-ject.

More generally, an arbitrary time varying gradientapplied after excitation causes transverse magnetization vec-tors to dephase according to the equation ,so that the total phase accumulation at timeis

(6)

Integrating the signal over the volume of interest in the sub-ject, we obtain

(7)

where is a constant for coil and electronics-related pa-rameters, and is the proton density. Now, introduce theso-called -space conjugate variables [10] by defining

(8)

and dropping the constant makes the Fourier transform(FT) relationship between data acquisition and image recon-struction for MRI clear, as follows:

(9)

A map of the spatial distribution of magnetization (i.e., animage) can be obtained from a map of the conjugate spa-tial domain . Equation (9) shows how MRI gradientpulse sequences can be designed to obtain sufficient datafrom -space locations for image reconstruction via an in-verse FT.

The availability of the fast FT (FFT) favors the collectionof data with a regular pattern in-space: e.g., a two-dimen-sional (2-D) Cartesian grid where, typically, one row of the

-space data matrix is acquired at a time, and the processis repeated over all such rows. Images are then reconstructedusing the 2-D FFT. Data collected in a polar sampling schemecan be reconstructed by a 1-D FT followed by filtered back-projection. More general acquisition schemes (e.g., spiralsampling) can be reconstructed by regridding the samplesonto a regular Cartesian grid and applying the 2-D FFT.

A pulse sequence diagram displays the timing of allthese gradients (see Fig. 1), RF pulses and acquisition of

Fig. 1. A pulse sequence diagram of spoiled gradient echoimaging displaying the timing of all major events during imaging.These are slice selective (Gslice), phase encoding (Gphase),and readout (Gread) gradients, applied RF pulses for excitation(RF), and acquisition (AQ) of induced signals, or echoes. Spatialencoding is for a 2-D Cartesian grid: Gread is on during theacquisition window; encoding in the orthogonal, phase encodingdirection is accomplished by repeating this acquisition step multipletimes with phase encode gradient pulses of increasing amplitudes.

induced signals, orechoes. Readers who wish to get a quickintroduction to MR scanner hardware and some furtherMRI terminology, and familiarize themselves with hardwarerequirements of cardiac MRI, are referred to [11]. In oneof the major modes of MR imaging,spoiled gradient echoimaging, a sequence of short low-amplitude RF (smallflip angle) slice-selective excitation pulses is played at arate of up to 400 pulses/s, and the MR signal is acquiredin an approximately 1–4-ms window after each RF pulse(see Fig. 1). Spatial encoding is achieved as for the 2-DCartesian grid given previously: a gradient is applied in theread direction during the acquisition window; encoding inthe orthogonal phase encoding direction is accomplishedby repeating this acquisition step multiple times with phaseencoding gradient pulses of increasing amplitudes used tooffset the data acquisition-space line.

The length of this pulse sequence repeat unit in Fig. 1 de-fines TR (repetition time), and the time from the center of theexcitation RF pulse to the center of the acquisition windowdefines TE (echo time). The contrast between the various tis-sues arises from the pulse sequence parameters (notably TRand TE), as well as tissue proton (i.e., nuclei) content,the tissue-specific magnetic relaxation properties, , and

.The expression for the MR signal obtained during a

spoiled gradient echo sequence, such as the one presentedin Fig. 1, is shown at the bottom of this page in (10), where

is the -space trajectory, typically covering arectangular grid, one line per TR, andis the RF flip angle.Note that the term

(11)

(10)

OZTURK et al.: ESTIMATING MOTION FROM MRI DATA 1629

Fig. 2. In segmentedk-space cardiac imaging data acquisition is synchronized, gated, with EKG.After each heartbeat same spatial encoding gradients for a specifick-space segment is acquiredrepeatedly, each is subsequently assigned to different images, providing multiple snapshots of cardiacactivity. The segments within each image are acquired in successive heartbeats. Please note that eachsegment may contain multiplek-space lines.

is the longitudinal steady-state magnetization resulting fromthe repeated flipping and spoiling of the magnetization,

represents the transverse component of the tippedmagnetization, and represents attenuation due tosignal decay during data acquisition.

It is the transverse magnetization that excites the currentin the receiver coil. The transverse magnetization has bothan amplitude (size of the vector in the transverse plane) anda phase (location of the vector in the plane). The phase of thetransverse magnetization can be manipulated by changingthe rate of precession of the magnetization with the appliedfield gradients. It is this information that is used to performthe spatial encoding to form the MR image. Each pixel inthe final MR image is in fact a complex number representingboth the magnitude and the phase of the transverse magneti-zation at time TE.

Neglecting relaxation effects, the signal-to-noise ratio(SNR) ratio for MRI is in proportion to the voxel volumeand the square root of the total acquisition time for theimage. In practice, the need to allow for some longitudinalrecovery of the magnetization may limit the minimum repeattime TR. Similarly, the effects of signal decay limitsthe duration of the acquisition readout window. Thus, MRIis essentially a constant optimization exercise with severalimaging parameters, preparation pulses, and gradients toget the desired image with adequate contrast and SNR in areasonable acquisition time.

Imaging the heart presents the additional challenge ofdealing with cardiac motion during the image acquisitionprocess. One approach to meet this challenge is to acquireconsecutive images as rapidly as possible, by using as short

a TR as possible. Images acquired in this fashion generallyhave limited spatial resolution (due to the limited numberof -space lines that can be acquired), and poor temporalresolution (150–300 ms). Improvements can be achievedby using echo-planar techniques (i.e., collecting several

-space lines per TR by modulating the gradients) [12]–[14]or by using longer -space readout trajectories [15], [16]in each TR. More recently, techniques known as parallelimaging have provided the possibility for accelerating theimaging process further via the use of multiple receiver coils[17], [18], albeit with some SNR penalty. The adoption ofthese techniques may make the direct acquisition of cardiacimages with high spatial and temporal resolution feasible inthe near future.

Segmented -space imaging is an alternative approachpermitting the acquisition of a set of images at multiple car-diac phases over the course of several heartbeats in a singleECG-gated, breathheld scan. This is achieved by partitioningthe -space data matrix into several so-called segments. Theregion corresponding to each-space segment is acquiredrepeatedly for the duration of a heartbeat, providing multiplecardiac phases, and the successive segments are acquired insuccessive heartbeats. The set of segments for each cardiacphase is then assembled and reconstructed into an image, asshown in Fig. 2. The temporal resolution can be improved byreducing the segment size (i.e., by reducing the number of

-space lines in a segment), but note that this is at the priceof an increased total imaging and, therefore, breath-holdtime. Segmented -space imaging techniques can be im-proved by using interpolation techniques to adjust for slightvariations in the cardiac cycle [19]. The acquisition times for


segmented-space imaging can also be reduced through theuse of echo-planar, spiral, and parallel imaging techniques.For further reading, several review articles exist for a generalintroduction to these different imaging techniques with afocus on their cardiac implementations [20], [21].

B. Overview of Different Techniques for Motion Tracking

The standard method of kinematics is to follow objectsusing temporal sequence of 2-D and three-dimensional(3-D) data sets. From these images, boundaries and edges ofdesired tissue is first identified by tissue segmentation tech-niques and subsequently tracked using object registrationmethods. Several imaging modalities can be utilized for thisdepending on the application. If the moving organ can beobserved directly (e.g., whole arm or foot), special markersfor landmarks, or active or passive stereo computer visiontechniques, can be employed. A rich literature already existsin this type of applications (e.g., gait analysis). Our focushere is imaging of moving tissues inside the body.

One of the methods forin vivo motion imaging is 3-DX-Ray stereophotogrammetry, which requires insertion ofmetallic balls into the bones [22]. This overcomes the lim-itations of working with 2-D radiographic projections butcannot be used as a routine tool. Computerized tomography(CT) can also be utilized [23], although X-ray dose becomescritical when multiple 3-D datasets are needed.

MRI has several advantages, since it provides excellentcontrast between soft tissues, and images can be acquired atpositions and orientations freely defined by the user withoutthe use of ionizing radiation. As an example, the 3-D motioncharacteristics of peritalar joint complex is analyzedin vivousing 3-D datasets acquired during the foot motion rangingfrom extreme pronation to extreme supination at eight posi-tions [24], [25]. Similarly, shoulder kinematics is examinedusing sequential increments of arm rotation and acquisitionand comparison of 3-D models of glenoid and humerus [26].

MRI can do much more than providing detailed anatomicimages as was the case in the previously mentioned kine-matics studies. Specific MRI techniques have the ability totrack tissue or give more quantitative motion data directly.These are: 1) TMRI; 2) PCMRI; and 3) pulse field gradient-based MRI methods (HARP and DENSE).

TMRI modulates the underlying image intensity with thehelp of a specific presaturation technique. This essentiallyproduces a pattern of dark lines, or so-calledtags, on theimage. Deformation of these temporary lines is analyzed toderive a motion model of the underlying tissue. In contrast,PCMRI uses phase shifts induced in the transverse magne-tization with “instantaneous” velocity encoding pulses formeasuring motion; the phase of the signal is directly relatedto the velocity of the material within each voxel at the time ofvelocity encoding. The velocity field can then be integratedto get the displacement. In DENSE and HARP, a uniformpattern of phase modulation is encoded into the tissue at achosen time, and the deformation of that pattern is detectedat a later time and utilized to estimate the motion.

This is a rapidly evolving area in MRI research; the latesttechnical refinements or detailed explanation of each tech-

Fig. 3. Left: standard cardiac MR images. Right: for the sameheart, location, and cardiac phase, the corresponding tagged MRimages. Top: at end-diastole, when the ventricular cavity is at itsmaximum filling, right after tagging. Middle: At mid-systole, whenthe heart is contracting. Bottom: at end-systole, when the maximumamount of blood has been ejected and the heart muscle is at itsmaximum contraction.

nique is not the scope of this paper. In the following subsec-tions we will introduce and describe each technique briefly,and their examples will be given in the next section.

C. Tagged Magnetic Resonance Imaging

In MRI, temporary magnetic fiducial markers, or tags, canbe created within the tissues. Subsequently, when the sametissue is imaged after a certain time, the shape changes ofthe tags reflect the underlying tissue motion (see Fig. 3). Theprinciple for tagging can be traced back to techniques for themeasurement of bulk flow [27]; over the past decade it hasbeen developed extensively for measuring cardiac motion.Several types of tagging approaches have been proposed inthe literature [29]–[32]. For detailed information on tagging,the reader is referred elsewhere [33], [34]. The parallel planestripe pattern and the combination of two orthogonal planetags forming a grid are the most common types. Basic tagsequences of this type are now integrated within the pulsesequence libraries of all clinical MRI machines.

The tagging operation may be considered to be a spatiallyselective excitation involving the combined use of RF pulsesand gradients. However, compared to slice selection (dis-cussed previously) in which only a single plane of magne-tization was excited, it is desired to excite multiple planes ofmagnetization for saturation tagging. The excited magnetiza-tion is then spoiled, i.e., dephased using gradient pulses, so


that it makes no significant contribution to the subsequentlyacquired images. The effect of the tagging excitation andspoiling is to leave null regions in the remaining longitudinalmagnetization that will appear as corresponding nulls in theimages at the tagged locations.

A number of preparation modules have been developed toperform the tagging operation. While some highly special-ized tagging methods have been developed to effect arbitraryspatial tagging patterns, we limit the discussion here to thegeneration of tagging patterns that comprise regularly spacedparallel lines or grids, and that use the FT approximation tomodel the effects of RF pulses.

The DANTE RF pulse scheme, which comprises a trainof very short RF pulses, is a useful basis upon which MRItagging procedures can be developed. An idealized DANTEpulse can be denoted mathematically as

(12)

for which FT theory shows that the frequency spectrum com-prises a comb with spacing . Practical imple-mentation of the sequence must employ a finite length combwith RF pulses of nonzero duration and finite amplitude

(13)

where denotes the shape of a single RF pulse andis the amplitude envelope of the comb. The response functionis now

(14)

where and are the FTs of and ,respectively. Hence, the width of the tagged comb willincrease as decreases and the shape of the excited tagsis controlled by . As with slice selection, the frequencyselective DANTE pulse is made spatially selective by thesimultaneous application of a gradient while the comb ofRF tagging pulses is being played.

Tagging using the DANTE pulse sequence [31], [35] iseffective on small (e.g., animal) imaging systems, where rel-atively short high-amplitude RF pulses can be achieved. Onlarger, whole body systems, however, the ability to play ar-bitrarily short but intense RF pulses is limited. For these sys-tems, it is more common to turn the gradient off during thetransmission of each RF pulse in the comb of RF pulses,yielding the SPAtial Modulation of Magnetization (SPAMM)tagging procedure [30]. This modification renders the indi-vidual RF excitations nonspatially selective, eliminating therequirement that the RF pulses be short to avoid excessivemodulation of the response by .

Between the RF pulses, the gradient pulse causes the mag-netization vectors to acquire phase in proportion to their po-sition in the tagging direction. Those vectors that acquire aphase of radians (for integer ) from each gradientpulse will experience the full, combined flipping effect of allthe RF pulses. However, for other vectors, the effect of the

Fig. 4. A typical SPAMM tagging pulse scheme to create a gridfrom two sets of parallel stripe tags at 45and 135 angles on theimaging plane. For each tagging direction, a series of RF pulsesthat are separated by tagging gradientblips are used. After eachset, a spoiler gradient is played to dephase the residual horizontalmagnetization. After these tagging pulses, standard imaging pulsesequences will follow.

RF pulses is generally much less coherent, and the combinedeffect over many RF pulses is negligible.

In practice, the tag spacing must be chosen such that tagpairs can be resolved by image processing, requiring 5–7-pixel separation between tags. Doubling the tagging densityto improve the spatial resolution of the motion data gener-ally requires a doubling in image resolution, which resultsin lower SNR. A method to overcome this is to perform theoriginal tagging procedure twice, with the tags shifted by halfthe tag spacing. Multiplication of these images together pro-vides an image with half the original tag spacing [36].

A typical SPAMM tagging pulse scheme is shown inFig. 4. The scheme to create a single set of parallel stripetags involves the application of 5 to 7 400-s RF pulses totag each spatial direction, sufficient to achieve a combinedflip angle of 90 –180 . The tagging gradient is blipped (i.e.,a quick gradient on and off) between each successive RFpulse in the comb; the time integral of each blip determinesthe inverse tagging separation. After the comb is played out,a large spoiler gradient is played in a direction orthogonalto the tags to dephase the excited fingers of magnetization.Grid tags are created by the successive creation of two setsof parallel stripe tags with tagging gradient pulses orientedorthogonal to each other.

Once created, the encoded tagging pattern decays overtime as the magnetization recovers by longitudinal relax-ation with time constant , as discussed previously. Themyocardium has a of approximately 850 ms, so thatpotentially, tag contrast can persist throughout an entirecardiac cycle. In practice, the imaging process furtheraccelerates this rate of tag fading. In each successive TR,a -flip angle RF excitation pulse samples the taggedcomponent of the magnetization, reducing it by a factor

: the available contrast is, therefore, rapidly depletedduring the early cardiac phases. In addition to depleting thetagging pattern, longitudinal relaxation causes regrowth ofthe untagged image term, producing an apparent gradualincrease in the signal at the tagged locations over the cardiaccycle, complicating the identification of the tags in latediastolic images. The complementary SPAMM (CSPAMM)


Fig. 5. The tag point displacementd, which is measured at thetag pointT at a given time, gives only one component of the pasttrajectory, so it cannot be mapped back uniquely. For example,pointT might originate fromp1, or fromp2. The correct matchingpoint p1 at the undeformed state (at tagging time) can only befound by incorporating the tag displacement information comingfrom other tag planes. Using all tags at a given time frame, we canlocate pointp1, but we cannot be sure of its trajectory (lines 2 and3 are two of the many possible paths). Using the matching pointsat every time frame, we calculate a final forward motion field andfind the correct trajectory.

method [37] provides an effective method to even out thesampling of the tagged magnetization, so that the taggingcontrast remains constant throughout the cardiac cycle. InCSPAMM, two successive 1–1 SPAMM imaging proceduresare performed with a phase cycling of the second taggingRF pulse. Subtraction of the two sets of images eliminatesthe relaxed/untagged term, providing tagged images inwhich the tags remain nulled throughout the cardiac cycle.CSPAMM also employs a scheme of ramped imaging flipangles [38] tuned to the tissue which are contrived tosample the tagged magnetization in such a way that thecontrast between the tags and myocardium is constantthroughout the cardiac cycle. Compared to nonramp-flippedimaging, tag contrast is reduced in the early systolic phasesand enhanced in the later diastolic phases.

The measured tag deformation at a single tag point con-tains only a unidirectional component of its past motion,from tagging to imaging time (see Fig. 5). In order to achievea full 3-D tracking of any point through time, the informa-tion coming from different tagging sets has to be combinedand interpolated in space and time. The classical analysisof TMRI contains three steps: 1) segmentation of the my-ocardium by drawing endo- and epicardial contours which isroutinely done interactively [39] or semiautomatically [40];2) detection of the tag points for each slice, tag orientation,and cardiac time frame; and 3) fitting a motion field (or tissuemodel) using three (or two for 2-D analysis) orthogonal 1-Ddisplacement information coming from all the tag detectedpoints.

Several approaches have been taken for the various steps inthe past. For the myocardial contour detection and tag anal-

ysis, we can name the approach of Guttmanet al. [40], [41]or the approach of Axelet al. [42], [43], which also includesfinite element-based cardiac modeling tools. Most of the dif-ferences in the various techniques come from the underlyingmotion model once the tag points are identified [44]. Someof them rely on specific geometry of the chambers and uti-lize a specific coordinate system: e.g., prolate spheroidal [45]or planospherical-based approaches [46] for the left ventricle(LV). Other techniques rely on finite element models. In thework of O’Donnell et al. [47], volumic superquadric withparametric offsets; in the approach of Partet al. [47], vo-lumic deformable superquadric; and in Young’s work [49],16-element cubic polynomials provide the model for the fi-nite element bases. In a completely different approach, themyocardial wall is decomposed into a finely spaced meshand displacement at each node is constrained via finite differ-ence analysis [50]. This method has been recently improvedto give unsupervised strain reconstruction on the LV [51].

B-spline-based methods have been proposed because theyprovide several attractive properties like parametric conti-nuity, compact representation of the information, local sup-port, and differentiability [52]. Several B-spline-based tech-niques have been employed in the past to describe deformedtags: Moultonet al. [53] utilized them to describe the de-formed tag planes as surfaces, but they used a global polyno-mial basis function expansion for the final forward tracking.In the first approach of Aminiet al. [54], tags were seg-mented using a 2-D coupled B-snake grid; they later ex-tended their approach to 3-D [55]. A step-by-step derivationof a B-spline-based four-dimensional (4-D) (3-D plus time)motion field [tagged tissue tracker (TTT)] is also proposedand evaluatedin vivo [56]–[58]. Examples of tag strain (E)analysis (TEA) and TTT analysis are given in Section III-A.

Most of the methods proposed in the literature for TMRI-based motion analysis separate tag detection and motion fieldfitting, but several combined approaches have also been pro-posed [59], [60]. The recently developed harmonic phaseimaging (HARP) technique has the potential of rapidly de-tecting tag locations without the need for myocardial seg-mentation [61]–[64]. HARP derives the motion informationfrom the noncentral spectral peaks in the frequency domain.When tagged images are Fourier transformed, the resulting

-space reveals multiple spectral peaks. When the first peakis isolated and inverse Fourier transformed, the phase con-tours in the resulting complex valued image behave simi-larly to a conventional tagging pattern, tracking the under-lying motion. At a user defined region, this approach canyield close to real-time strain analysis. A myocardial seg-mentation method with a high degree of automation has alsobeen proposed for HARP making a more automated myocar-dial motion analysis feasible [65]. The HARP approach canalso be used to acquire faster motion encoded images by tai-loring the MR pulse sequence. The acquisition approach ofHARP (FastHARP) will be covered more in Section II-F.

D. Phase Contrast Magnetic Resonance Imaging

A different approach to motion analysis is based on thesensitivity of the phase of the MR signal to motion. It was


Fig. 6. Basic pulse sequence diagram unit of a conventionalPCMRI. For each line ofk-space acquired, two scans are acquired.Velocity encoding gradient in the phase encoding direction isindicated here by the solid bipolar gradient pair.

intended initially for blood flow measurements [65]–[67],but it is now used to obtain strain measurements of the my-ocardium [68]–[73].

This technique gives pixel-by-pixel value for the under-lying tissue’s velocity. The basic principle is to acquire twodatasets with two different velocity encoding gradients butotherwise identical acquisition parameters and to subtract thetwo phase images (see Fig. 6). The resulting difference imagewill be proportional to the flow (or tissue motion) if the solu-tion (or underlying tissue) can be assumed to have a constantvelocity during the acquisition window.

Velocity encoding gradients are bipolar, so they do not af-fect stationary protons but impart phase shifts to moving pro-tons. The equation describing the magnetization at TE forthis pulse sequence is

(15)

where is the velocity sensitivity of the velocity en-coding bipolar gradient pulse.

In order to eliminate the phase effects of sources other thanflow or motion, a reference scan is acquired. Since velocityinformation can be obtained only in one direction at a time,four independent measurements should be obtained to ob-tain a 3-D dataset. Whether imaging flow or tissue motion,careful planning of the velocity encoding gradients is nec-essary to eliminate aliasing or unintentional signal cancella-tion.

For deriving myocardial strain in PCMRI, as with the caseof TMRI, there are several different numerical approaches.Van Wedeenet al. [68] calculated the strain rate tensor di-rectly from the velocity data. Zhuet al. [71], [72] obtained adisplacement field from the velocity data and then used thisfield to calculate the strain tensor.

E. Introduction to Pulsed Field Gradient Methods

The inherent motion sensitivity of MR was recognizedshortly after the discovery of the MR phenomenon [74]–[76].Remarkably, proposals employing the use of magnetic fieldgradients and FT methods for the measurement of velocitydistributions [77], [78] were made even before the advent ofMR imaging. Since that time, many techniques have been de-veloped employing a common motion encoding mechanism:

a pair of pulsed field gradients (PFGs). A major advantageof the PFG schemes is that they provide a direct measure ofthe tissue displacement, greatly simplifying image postpro-cessing requirements compared to either time of flight andinflow enhancement methods.

Referring to Fig. 7, the first pulsed field gradient ofamplitude and duration causes the magnetization vec-tors to rotate with an additional frequency proportional totheir instantaneous location, causing a phase accumulation

. After an evolution period during whichthe tissue can move to a new location , a secondgradient pulse, logically opposite to the first, is applied.After the second pulse, the net phase accumulation is

, thus encoding the displacementthat occurred during the motion evolution interval in thephase of the magnetization. By combining PFG methodswith MR imaging sequences, maps of displacement can bedirectly obtained: the displacement is proportional to thephase of the complex-valued sample at each image pixel.The precision of the motion measurement obtainable withPFG methods is determined by the areaof the pulsedgradients, and is independent of the spatial resolution ofthe image. However, it should be noted that the measuredvalue represents a superposition of all the signals fromthe whole voxel, and so is an average of the motion takenover the whole voxel. Also, since the motion informationis encoded as phase, the possibility of aliasing must beconsidered either by restricting the range of phase evolutionto , or by obtaining additional measurements toassist in resolving the ambiguity.

The phase contrast velocity encoding mechanism (de-scribed previously) can be viewed as a PFG technique inwhich the assumption of uniform velocity during the motionevolution period (which is kept small) is made. More gen-erally, where longer motion evolution periods are employed,uniform, bulk displacements of the magnetization producesignal phase shifts in proportion to the PFG area and interval

. Simultaneously, dispersive motions (diffusion, sheer,compression, and rotations) occurring in a given voxel willcause signal attenuation due to destructive interferencecaused by the reduction in phase coherence [77].

In practice, variants of the simple PFG pulse sequence aregenerally preferred. The bipolar encoding sequence is notwell-suited to cardiac motion encoding because the signal:1) decays rapidly with and 2) suffers phase distortionsdue to local main field inhomogeneities . The intro-duction of a spin echo yields the pulsed gradient spin-echo(PGSE) sequence, reducing attention to the effects ofandrefocuses inhomogeneities. The pulsed gradient stimulatedecho (PGSTE) sequence [79] reduces attention tobystoring one component of the magnetization along,permitting even longer motion encoding periods. Recently,methods known as DENSE [80] and HARP [81] have beenproposed for imaging myocardial function, which employ astimulated echo encoding technique.

A single motion encoded sample generally does not pro-vide sufficient information to recover the actual displace-ments due to the presence of additional sources (e.g.,


Fig. 7. A simple PFG sequence, where two opposing gradients which are separated by a certainduration causes the magnetization vectors to first rotate and later return to their original phase valuefor stationary protons, causing a net phase accumulation only for nonstationary protons. Twovariants of this simple PFG pulse sequence are also given: pulsed gradient spin-echo (PGSE)sequence with the introduction of a spin echo and the pulsed gradient stimulated echo (PGSTE)sequence, which stores one component of the magnetization along longitudinal magnetization (seeSection II-E for details).

Fig. 8. The basic pulse sequence unit of DENSE, which employs a cardiac and respiratorygated PGSTE encoding scheme. Four values of the motion encoding gradient are needed toprovide a reference and sensitivity to motion in all three spatial directions. For FastDENSE andMetaDENSE, see Section II-F.

inhomogeneity, RF coil phase) of phase in the MRI signal.Fortunately, these spurious effects can largely be canceledthough the use of reference scans (in which different PFGsare employed) and phase cycling schemes. These sensitivi-ties permit the use of PFG methods to obtain measures ofdisplacements and evaluate derivative measurements for ve-locity, strain, and elasticity.

F. DENSE: Displacement Encoding With Stimulated Echoes

Aletras et al. [80], [82], [83] have recently proposed afamily of highly sensitive motion-encoding schemes, collec-tively termed DENSE, to obtain cardiac strain maps. TheDENSE methods are implementations of the PFG motionencoding sequence, employing a cardiac gated PGSTE en-coding scheme with increasingly complex strategies to re-duce imaging time and eliminate spurious sources of artifactand error.

The original proposal [80] comprises a direct implementa-tion of the PFG sequence with respiratory and cardiac gatedgradient echo imaging (see Fig. 8). Four values of the mo-tion encoding gradient were employed to provide a referenceand sensitivity to motion in all three spatial directions in ascan time of approximately 20 min. While inefficient froman imaging standpoint, just one-space line was collectedper respiratory cycle, and the sequence demonstrated proofof concept for the technique yielding convincing maps ofmyocardial displacement and strain. Subsequent versions ofDENSE, known as FastDENSE [82] and MetaDENSE [83],

address the inefficiency of the data sampling strategy to makeDENSE methods more clinically relevant.

FastDENSE employs a STEAM [84], [85] imagingmethod with an echotrain readout that repeatedly samplesthe PFG encoded magnetization during a relatively quietpoint in the cardiac cycle to improve acquisition efficiency.FastDENSE permits in-plane displacement maps and re-lated strain maps to be obtained in a single, 24-heartbeatbreath-hold time duration.

G. FastHARP: An MR Acquisition Scheme Optimized forHARP Analysis

The HARP technique [61]–[64], introduced in Sec-tion II-C, was originally proposed as a postprocessingmethod to quantify in-plane motion from tagged MRimages. The HARP method involved the acquisition ofconventional SPAMM tagged images from which a singlespectral peak corresponding to the fundamental taggingspatial frequency was isolated by band-pass filtering. TheFastHARP pulse sequence [81] is intended to directlyacquire the small region of-space data required for HARPprocessing rather than acquiring a full-space dataset.

Tagging in FastHARP is performed using the simplesttagging sequence, 1–1 SPAMM, which comprises just two90 RF pulses. The 1–1 SPAMM sequence generates asinusoidal tagging pattern corresponding to two taggingspectral peaks offset symmetrically from the-space origin.Generally it is desirable to generate as few tagging spectral


Fig. 9. A schematic of the FastHARP pulse sequence. It starts as a 1–1 SPAMM tagging pulse,followed by a gradient, which is needed to center thek-space on one of the tagging spectral peaks. Itends with an echo-planar type of acquisition to further speed things up. If the acquisition is gatedand the wholek-space is acquired in a single shot, repeating this sequence unitn times within acardiac cycle, would yieldn cardiac frames.

peaks as possible in order to maximize the SNR of thesubsequent FastHARP images. This simple tagging schemeis efficient because the available signal energy is dividedbetween only two tagging spectral peaks, maximizing thesignal that is collected. Data acquisition is offset in-spaceso that only a small region centered on one of these taggingspectral peaks is acquired.

A schematic of the FastHARP pulse sequence is shown inFig. 9. The FastHARP sequence can now be related to a stim-ulated echo PFG motion-encoding scheme (PGSTE). The1–1 SPAMM tagging preparation is identical to the initialPFG premotion evolution encoding pulses. For FastHARP,the data acquisition must be offset in-space to center it onone of the tagging spectral peaks. In practice, this is achievedby modifying the prereadout dephaser gradient, by addinga gradient area to cause the desired-space offset. The ad-ditional area required is precisely equal to that used in theSPAMM tagging procedure, i.e., the first PFG gradient pulse.Thus, the prereadout gradient pulse is a combination of astandard imaging read dephaser plus the second PFG mo-tion encoding gradient pulse. It is clear that the stimulatedecho PFG and FastHARP employ essentially identical phys-ical principles in their motion encoding mechanisms.

The method, however, faces a number of challenges. TheSNR of the HARP images is limited because the availableencoded magnetization must be divided among the mul-tiple cardiac phases that are to be imaged in each cardiaccycle. Also, the regrowth of longitudinal magnetization byrelaxation is a source of confounding signals in FastHARPimages. Recently, an imaging scheme employing rampedflip angles and CSPAMM-like phase cycling of the secondSPAMM RF pulse (albeit with a doubling of the effectiveimaging time) has been demonstrated [86] to address theseissues. More generally, to produce datasets without breathholding, imaging must be completed for each encodingdirection in just one to two heartbeats. This limits the degreeof segmentation of -space that can be performed, which inturn limits the size of the data matrix that can be acquiredfor a given temporal resolution, and encourages the use ofhighly efficient imaging methods.

Since the data processing for PFG methods is straightfor-ward and does not require operator attention, FastHARP po-tentially provides a method for real-time strain monitoring ofcardiac function without patient breath holding.

III. A PPLICATION EXAMPLES

Here we present several applications of motion analysistechniques. We will mainly focus on cardiac implementa-tions, finishing with an example of tongue motion analysisat the end.

A. Regional Myocardial Motion Analysis Using TMRI

MR tagging is the only imaging modality to date that hasbeen able to quantify local contraction over the entire heart[87], [88]. A database of systolic 3-D strain patterns has beendeveloped for the LV in healthy subjects [89]. This is a veryactive research area; detailed mechanical analysis of the LVhas been performed to quantify changes due to ischemia [90]or pressure overload [91], to describe the effects of or re-covery from surgical procedures [92], [93] and effects ofdrug treatments [94]. We will start next with two quick ex-amples of LV strain analysis and later present a more detailedexample of a detailed biventricular motion analysis.

Our first example for tag motion analysis is detailed strainanalysis over a canine LV in various pacing conditions in acanine model [95]. Tagged MR images are analyzed withthe help of a LV-specific method TEA [45] after obtainingthe tags using the Findtags program [41]. Strain images ina single canine heart for two pacing conditions are shownin Fig. 10. When the ventricle is paced from the right ven-tricle (RV) free wall, the contraction begins at the pacing siteand spreads around the LV. There is significant prestretch onthe heart wall opposite to the pacing site. When the heart ispaced from the LV free wall, the pattern is the same, but theorigin of the contraction wave is moved to the new pacingsite (not shown). When paced simultaneously at both RVand LV pacing sides, resulting electrical excitation producesa more coherent contraction pattern both in space and time(see Fig. 10). This pattern is similar to normal physiologywhen the electrical signal originates at the right atrium and


Fig. 10. Strain images in a single canine heart for three pacingconditions. Contraction is shown as blue, stretching is yellow, andpacing site is shown with an asterisk. Evolution of strain over theLV is shown for atrium (physiological) and two ventricle-originatedpacing protocols.

Fig. 11. Comparison of myocardial strain images from a normalLV (top row) and an LV of a patient with dilated cardiomyopathy(bottom row).

spreads concurrently to both ventricles through a specializedconduction system.

An important research area is devising custom designedpacing protocols for ischemic and dilated cardiac diseases[96]. In Fig. 11, a comparison of myocardial strain imagesfrom a normal and a dilated heart (dilated cardiomyopathy)is made. In the patient, an early contraction of the septum(9 o’clock) is evident, with a late contraction of the LV freewall (3 o’clock). Techniques such as these can be used toevaluate the remodeling of the heart after the onset of is-chemic pathologies or during pacing [97].

We will now present activation time analysis using thelocal time-strain relationship. Data presented in this sectionwere obtained from MRI recordings of a canine heart,which was paced from the atrium or the right ventricle(RV) by using a GE Signa 1.5 T scanner with segmented

-space acquisitions with breath-hold periods. The scanningparameters used were a 28–32-cm field of view, time torepetition of 6.5 ms, time to echo of 2.1 ms, 25696 ac-quisition matrix, 32-kHz bandwidth, two to three readoutsper movie frame, in-plane spatial resolution of 1.253 mm,and slice thickness of 6–7 mm. We employed a motion anal-ysis method based on a 4-D B-Spline tensor representation[56], [57]. Local deformation gradients were obtained by

using this parametric model, and 3-D strain values wereestimated by using finite strain tensor analysis computedover a mesh located at mid myocardium of the RVs andLVs. The motion field and activation time calculations aredone automatically once the tags have been detected. Twoventricles are modeled differently; the LV was modeledas a deformed cylinder, and the RV as a deformed halfcylinder. A mesh is defined on these cylinders, and fromthe motion analysis, we obtain the time series of strain oneach mesh point (see Fig. 12). These graphs are subsampleddramatically in both directions for display purposes. Wehave opted to use circumferential strain (Ecc) values in ouranalysis because the orientation of the major muscle fibersin mid-myocardium are known to be in that direction [98].

The activation time has been chosen as the time of peakEcc strain value. When no peak is present, neighborhood-based correlation is employed as in [95]. The calculated ac-tivation time values are further scanned for local activationwave direction inconsistencies to eliminate velocity compu-tation at sink or source locations. RV and LV analyses aredone independently.

If we compare the time-strain plots for both pacing condi-tions [see Fig. 12(a)–(d)], for atrial pacing, we obtain similarlocal strain time graphs over the LV and RV. Strains decreasealmost linearly during ventricular systole. On the other hand,for ventricular pacing, we observe double peaks due to ini-tial or delayed stretch; definitely a more complex time-strainpattern is observed with the start of contraction delayed atcertain locations.

Another helpful display method for showing the activa-tion times over both ventricles is the bulls-eye graph [seeFig. 12(e) and (f)]. In this graph, apex of the LV is the centralring, and the basal part is the most outer ring. RV is shownon the right side with partial rings; color represents the acti-vation time from a reference time in imaging. It provides aquick display of the overall activation pattern of the heart.

In bulls-eye plots, for atrial activation, we see relativelysmall activation time range (134 ms) over both ventricles.Mechanical activation seems to start at several locationsat once and proceeds in various directions. Some of thebasal portion of the LV is activated last. These findingsare compatible with the known physiological excitationpatterns through the Purkinje system. This is also supportedmostly by higher propagation speed than ventricular pacing(not shown). However, we should note that a definitedistinction between propagation over the Purkinje systemor myocardium itself cannot be made definitely using thisdata. On the ventricular paced heart, clearly, a propagationwave can be identified, starting from the RV and proceedingtoward the LV. This, and slower speed of activation wavepropagation, results in a higher range of the activation timesfor the LV (206 ms).

B. Ventricular Blood Flow Analysis Using PCMRI

Our PCMRI examples come from the work of Thompsonet al. [99]. In this study, a high temporal resolution PC tech-nique is employed. Improvement by a factor of two in the


Fig. 12. Strain values over the LV during systole of a canine heart for: (a) atrial excitation and (b)RV excitation.Y -axis represents strain;�0:1 means 10% shortening; positive values representstretching. Calculated activation times for each location are displayed on the small strain evolutiongraphs with a small vertical line. The small plots in each row represents mesh points arrangedcircularly in a short axis cut of the heart. Coming down on each column represents moving from thebase of the heart toward its apex. Strain values over the RV during systole of a canine heart andcalculated activation times are displayed in the middle again for: (c) atrial excitation and (d) RVexcitation. (e) and (f) Corresponding biventricular bulls-eye plots of activation times are shown.Note that the widened activation time scale (in milliseconds) and propagation of activation fromright to left for RV pacing [see Fig. 12(f)].

temporal resolution was achieved by acquiring the differen-tially flow-encoded images in separate breath-holds. Addi-

tionally, a multiecho readout was incorporated into the PCexperiment to acquire more views per unit time than is pos-


Fig. 13. Cardiac velocity images in all three directions at endsystole. Top left:V x, velocity in the horizontal direction. Top right:V y. Bottom left:V z. Bottom right: magnitude image. The in-planeresolution is 2.0 mm� 2.0 mm for both the magnitude imageand all the vector images. (Images are courtesy of R. Thompson,National Institutes of Health, National Heart, Lung, and BloodInstitute, Bethesda, MD).

sible with the single gradient-echo technique. A complete setof images containing velocity data in all three directions wasacquired in four breath-holds on normal volunteers, with atemporal resolution of 11.2 ms and an in-plane spatial res-olution of 2 mm 2 mm. For the measurement of all threevelocity components, phase-difference images were calcu-lated using velocity encoded data. Sample velocity imagesin three logical axes ( , , ) and a magnitude imageon a mid-cardiac short axis plane are shown in Fig. 13. Hori-zontal and vertical gradients can be observed forandimages over the myocardial wall of the LV, since it predomi-nantly contracts toward its center. On the other hand, there isa uniform through-plane motion of the myocardium for thisslice, which is displayed nicely on the image.

The finer details of myocardial motion, like twisting, re-quire combined vector analysis of the velocity images. Forexample, vector analysis at this high frame rate can also beused to analyze fine details of blood motion patterns (seeFig. 14).

C. DENSE and Cardiac Motion Analysis

To date, DENSE methods have concentrated on obtaininghigh precision and high-resolution measurements. To maxi-mize the available SNR and permit high-resolution imaging,the sequences acquire an image only at one phase of the car-diac cycle. Strain measurements are then obtained for thetotal systolic contraction.

FastDENSE acquisitions have been initially reported witha 128 96 acquisition matrix, 2.5-mm spatial resolution,and acquiring a single cardiac phase (26-ms temporalresolution) with a total imaging (i.e., breath-hold) time of24 heartbeats [82]. The MetaDENSE methodology reducedthe breath-holding time for a similarly specified scan to

Fig. 14. Left: a single three-chamber image (frame 59) from100 frames reconstructed from a multiecho, multibreath-hold PCacquisition from a healthy volunteer. Right: a velocity vector plotcalculated from the in-plane velocities, from the image plane shownon the left, for a 32-� 60-mm rectangular region of interest. Thein-plane resolution is 2.0� 2.0 mm for both the magnitude imageand the vector field image. LA = left atrium; Ao = aorta; LV =left ventricle; AMVL = anterior mitral valve leaflet. (Images arecourtesy of R. Thompson, National Institutes of Health, NationalHeart, Lung, and Blood Institute, Bethesda, MD).

Fig. 15. Sample meta-DENSE-based 2-D strain analysisresults from a normal volunteer overlaid on the corresponding amid-ventricular short axis slice. The superimposed lines representthe direction of the principle strains with colors representing theireigenvalues. (Images are courtesy of A. Aletras, National Institutesof Health, National Heart, Lung, and Blood Institute, Bethesda,MD).

just 12 heartbeats. MetaDENSE [83] improves the imageacquisition efficiency, by utilizing a train of 180refocusingpulses resembling a fast-spin-echo (FSE) readout, instead ofthe STEAM-EPI-based scheme. The FSE readout employedin DENSE is long ( 100 ms), and so must be performedduring a quiescent period of the heart cycle. In addition, thesequence incorporates several features to reduce artifactsand associated errors in the displacement encoded data. Thesequence employs a 180inverting RF pulse approximatelyhalf way through the motion encoding interval in order toeffectively nullify the (nonmotion encoded) signals fromtissue that relaxed during the evolution period. In effect, thisinverting RF pulse serves the same purpose as the RF pulsephase-cycling scheme in CSPAMM without doubling theimage acquisition time, but can only nullify the signal for aparticular T1 species (e.g., myocardium), and cardiac phase.

Fig. 15 shows the results of a typical MetaDENSE studyof a mid-ventricular short axis slice from a normal volunteer.The superimposed vectors represent the direction of


Fig. 16. (Top) Nonbreathheld and (middle) breathheld examples of fastHARP acquisition. Bothimage sets are acquired in two heartbeats. These images have been reconstructed from originaldata with synthetic tags to resemble (bottom) the equivalent tagged image data set. They are alsocolored displaying contraction (blue) and relaxation patterns all around the ventricle wall in a normalvolunteer. (Images are courtesy of S. Sampath, Johns Hopkins University, Baltimore, MD).


Fig. 17. Real-time, tagged MRI images of the tongue during thesyllables “sha” at 164, 350, and 600 ms. Intersection of the tagswith the tongue surface are also displayed at each time frame.

principal strains, which corresponds to local circumferentialand radial strains. These two different types of strain pat-terns can be identified by the negative and positive eigen-values, which are used in the color-coding of line segmentsin Fig. 15. In a normal myocardium during systole, a uni-form circumferential shortening is observed due to myocar-dial fiber orientation. The myocardial volume is kept almostconstant; therefore, circumferential shortening is accompa-nied by a similarly uniform radial thickening in normal indi-viduals.

D. Example HARP Analysis for Cardiac Motion

Development of FastHARP has been focused towardrapid imaging for the real-time monitoring during cardiacstress function testing where conventionally tagged imagingsuffers from several drawbacks due to the requirement forrepeated breath-holding and the lengthy time required forimage postprocessing. FastHARP currently employs aninterleaved, 8-echo echotrain (EPI) acquisition to collect a32 32 matrix at 9 mm spatial resolution. Although thisresolution is low compared to conventional anatomicalimaging, it is comparable to the resolution of the motiondata (i.e., the tag spacing) in a conventionally tagged image.

In each heartbeat, the sequence yields 12–20 cardiacphases (46-ms temporal resolution) of motion-encoded data,with one in-plane spatial direction being encoded. Twoheartbeats are required for complete displacement and strainmapping, because, as with other PFG methods, referencescans are required to suppress spurious sources of phasein the images. These can be either acquired at the begin-ning of scanning, or interleaved with the real-time cardiacstrain mapping. Nonbreathheld and breathheld examples ofacquisition are given in Fig. 16 along with correspondingstandard SPAMM images for comparison. The displayedFastHARP images have been reconstructed from originaldata with synthetic tags to make comparison easier.

E. Motion Analysis of the Tongue

As an example of the motion analysis of a noncardiactissue, we provide here an example of the analysis of tonguemotion. Such an analysis can give important clues aboutspeech control and causes of various speech pathologies.

Imaging the movement of the tongue is challenging becausethe selected technique must image without interferingnormal tongue motion and must be capable of recording itsrapid motion. Different research groups have used differenttechniques, including X-ray [100], X-ray micro-beam [101],electropalatography [102], electromagnetic articulography[103], electromyography [104], [105], ultrasound (US)[106], and MRI [107]–[109], the MRI being our currentfocus.

For the speech material a consonant-to-vowel syllable(“sha”) was chosen due to its large tongue motions and shortdurations (see Fig. 17). Here, a modified gated multiechoSPGR sequence in a 1.5-T GE Signa cardiac scanner isused (TR 11.3 ms, TE 1.5 ms, FA 10, ETL 8, NVP 40, BW125, 128 96 matrix, NEX 0.8, FOV 30 15 cm, ST 7mm, tag spacing 5 mm). For these images, only one syllablerepetition per slice is used, in contrast to the work of Dicketal. [110] where a segmented-space acquisition was used.A gating input was provided to the scanner in sync with thecue. During 1 s, 16 images are acquired for a given slice.The acquisition is repeated for all the slices. For the full4-D (space and time) displacement and strain analysis [57],three sagittal and eight axial slices, with grid tagging, areobtained. Previously, a preliminary 2-D surface deformationanalysis was done for all four syllables [111]; here wepresent the first 4-D tongue motion analysis as an exampleof noncardiac tissue motion tracking. Fig. 18 displays a 4-Dcompression-expansion analysis of the internal points of thetongue for tagged MR images during the utterance of thesyllable “sha.”

IV. DISCUSSION

It is known for a long time that myocardial ischemia dueto lack of adequate blood supply to the heart muscle causesmyocardial wall motion abnormalities [112], [113]. Actually,local motion abnormalities are shown to be one of the ear-liest signs of the ischemic disease, when most of the globalmeasures are kept normal by compensating mechanisms inother regions of the heart [114], [115]. Precise and accuratemethods for measuring myocardial motion can be helpful inthe early diagnosis and accurate staging of diseases and quan-titative evaluation of new treatment modalities.

In this paper, we reviewed several techniques that areunique to MRI to measure precise tissue motion, with aspecial emphasis on the heart. Various imaging techniquesallow the acquisition of dynamic sequences of 3-D (4-D)heart images: various nuclear medicine techniques (likegated SPECT), X-ray CT, and US. There are several advan-tages of MRI over others: high spatial resolution, excellentsoft tissue contrast, intrinsic 3-D acquisition, ability toacquire images from any arbitrary plane, and absence ofionizing radiation.

Although we have focused on unique methods for MRI,classical motion analysis approaches can also be applied toMR derived images. These techniques utilize 4-D datasets(surfaces or volumes) and are essentially similar amongall imaging modalities (e.g., CT, US). Due to high tissue


Fig. 18. Four-dimensional compression-expansion analysis of the tongue during the utteranceof syllable “sha.” Top row: we observe two compression regions in sagittal view at the first timeframe: one in the upper-center and the other in the bottom-left part of the tongue. The compressionregion seen at the first time frame disappears later and we observe a huge compression region at theback of the tongue in the end. Bottom row: for the axial images, we observe compression in themiddle-back and expansion in the front regions of the tongue in the first time frame. Compression inthe middle-back region observed in “sha” at the first time frame, moves to the back of the tongue anddisappears in the last time frames. At the end of the utterance “sha,” expansion is observed in the tip.

contrast and resolution, which is significantly improved withthe reintroduction of balanced gradient imaging techniques[116], such approaches are usually much easier to implementfor MR. Historically, cardiac motion analysis, especially ofthe LV, has been the main focus of such approaches. Genericdeformable surface models [117]–[119], models exploitingcurvature information [120]–[123], 4-D models [124], ormodels with temporal constraints [125], [126] have beenpreviously described for cardiac motion analysis, and theyare all based on surface tracking.

The major shortcomings of surface-based approaches arethe inability to display transmyocardial strain differences andthe difficulty in describing the rotational component of car-diac motion. The clockwise and counterclockwise twistingof the LV during contraction and relaxation has received sig-nificant attention in recent studies, and disturbances in thesecomplex motion parameters are being investigated as earlysigns in various pathologies [127], [128].

The methods that are described in this paper (TMRI,PCMRI, DENSE, and HARP) are unique to MRI becausethey allow the tissue motion to be measured more directly. Asummary and side-by-side comparison of these techniquesis presented in Table 1. In TMRI, temporary markers are putwithin the tissue and then followed over time. In PCMRI,DENSE, and PCMRI, we essentially get a velocity ordisplacement image. These are in sharp contrast with theinferred motion, which is derived from surface or volumematching techniques.

Of course, the surface-based tracking techniques andmethods based on TMRI (or other MR specific methods) arenot mutually exclusive. For the tag-based motion analysisof the whole heart, a combined approach is needed, becausetagging is not possible on the relatively thin walls of theatria. Even for the ventricles, using both techniques togetherin combined optimization has been proposed but has notbeen fully developed [129].

Detailed mechanical analysis of the heart is certainly animportant research tool. Tagged MRI has shown to producemechanical activation maps that are closely correlated withthe corresponding electrical activation maps for the LV

[130], [131]. Generating electrical activation times requiresan invasive approach and can be obtained only on the en-docardial surface using specialized catheters, or accessiblepericardial surfaces. As explained in Section III-A (alsoin Figs. 10–12), a detailed mechanical analysis can beperformed for both ventricles, and combined mechanicalactivation maps can be created [132], [133].

Comparative analysis between these different approachesof motion tracking in MR is rare in the MR literature, be-cause most of the methods remained experimental and re-stricted to certain labs (and to certain type of scanners). De-clercket al. [44] proposed a platform for the comparison ofthe different techniques for TMRI and did a detailed compar-ison of four approaches: 1) TEA [45]; 2) discrete model-free(DMF) method [50]; 3) 4-D tag strain (E) analysis (FTEA)[46]; and 4) TTT [57]. All of these methods need the seg-mented tag points as inputs and provide different approachesto motion field fitting: continuous (TEA, FTEA, TTT) ordiscrete (DMF) in space, using LV specific (TEA, FTEA)or the standard Cartesian coordinate system (TTT, DMF).These methods have been compared using synthetic, exper-imental animal and human data, and their performance hasbeen quantified. They conclude that the choice of a specificmethod has to be based on the individual requirements fornoise sensitivity and spatial resolution, which has been quan-titatively classified. For example: DMF uses the least amountof smoothing; therefore, it has the highest spatial resolution,but the highest noise sensitivity. FTEA uses more smoothingthan the other three and is the least noise sensitive, but it hasthe lowest spatial resolution. Despite these small differences,they concluded that all four methods produce strain maps thatare useful for characterizing myocardial function with greatprecision [44].

Recently, Aletraset al. [134] compared the results ofTMRI with DENSE in six normal volunteers. Displacementand circumferential shortening measurements were found tobe very similar between the two methods. On the other hand,radial thickening measurements of the LV were significantlydifferent. He suggested that the overestimation in TMRIwas possibly due to the undersampling of motion along the


Table 1A Quick Comparison of Different Motion Analysis Techniques That Are Unique to MRI

radial direction. A disadvantage of the tagging techniquesis the fact that the spatial resolution of the functionalmyocardium compared to the anatomic resolution of theunderlying images is reduced by a factor of three to five inlinear dimensions due to the need to be able to resolve thetags in the images. On the other hand, tagging can providea time series of images through the cardiac cycle, whichmay aid in the identification of abnormal contraction. TheDENSE methods currently provide only a single measure ofoverall systolic contraction. Further comparative research iscertainly needed among different techniques.

Although FastHARP was developed based on the frame-work of magnitude tagging, further analysis reveals that itis, in effect, a variant of the PFG technique. The initial pre-motion-evolution encoding schemes are identical: an initialnonselective excitation, subsequent encoding with a pulsedgradient, which is followed by storage of the encoded mag-netization by a second nonselective RF pulse (see Fig. 9).After the motion evolution interval, the stored magnetiza-tion is reexcited with a slice selective RF pulse. The offset

-space data acquisition is achieved by modifying the pre-readout dephaser gradient pulse, adding gradient area equalto that used in the tagging preparation gradient pulse. Thisadditional pulse area is, effectively, the second PFG pulse.

With the improvements of PCMRI techniques [99], it isnow possible to achieve adequate resolution appropriate forimaging blood flow within the cardiac chambers, in addition

to conventional vessel flow imaging. In the past, such flowpatterns due to valve opening and closures, and indirectlypressure gradients, could be examined only by Doppler tech-niques. PCMRI with both high temporal and spatial resolu-tion can offer some advantages; e.g., it has neither an acousticwindow nor a source angle variability problem.

The heart motion is periodical and is well suited for gatedMRI techniques. Other moving tissues such as the tongueprovide additional challenges for MRI. For example, humantongue has a complicated muscular structure that is not easilyexplained by the deformation patterns seen on the surfaceof the tongue during speech. MRI-based motion analysis isshown to provide an important insight into the internal dy-namics of the tongue [135], and in Section II-G we provideda preliminary sample result of a 4-D compression analysisof tongue during utterance of “sha” using tagged, real-timeMRI.

Other noncardiac applications of tagged MRI is certainlypossible; the limitation comes mostly from the tag fadingand the difficulties with synchronization of the motionwith image acquisition, which requires a special hardwarefor each application. An interesting example is presentedby Huszaret al. [136], where they quantified the chestwall deformation during cardiopulmonary resuscitation(CPR) using tagged MRI. The development of effectiveCPR machines or techniques, or understanding the effectsof external trauma on internal organs, can be effectively


examined by the techniques presented here. Lung motion isalso examined during normal respiration using MRI tagging[137]. When the flow is relatively slow and uniform, as it isthe case for the cerebrospinal fluid, MRI tagging can also beapplied directly to quantify fluid motion [138]–[140].

Another set of applications of dynamic MR imaginginvolves detailed evaluation of certain internal tissue move-ments that can occur physiologically or under pathologicalconditions. In these applications, the subject is asked toperform certain maneuvers (e.g., straining) while insidethe magnet, and the newly acquired images are comparedwith the control images. A common target of this type ofapplication is the assessment of pelvic floor defects [141],[142]. Although the initial aim is to confirm the initialdiagnosis and evaluate the exact anatomic defect of deeperstructures, detailed quantitative analysis using the advancedtechniques in this paper can be performed.

Standard 1.5-T closed MR scanners have proven theirvalue in the evaluation of normal tissue anatomy and itsintegrity in the case of general trauma [143]. Dynamic MRimaging of the musculoskeletal system is usually performedin dedicated or open MR systems. With the advances ofimaging technologies, low-field MRI systems claim nowspecificity and sensitivity that are equal to or better thanhigh-field systems in identifying many musculoskeletalpathologies, e.g., identifying and characterizing meniscaltears [144]. These low-field systems allow a wide range ofdynamic studies; these are routinely analyzed only visuallyor using surface-based motion-tracking techniques.

In musculoskeletal imaging, detailed kinematical modelscan be built using MRI-derived 3-D structures. Exampleapplications are the analysis of the motion of the spine [145],patella [146], wrist [147], [148], ankle [24], [25], shoulder[26], or building realistic models for general locomotion[149]. Although, these are extremely valuable as a researchtool, such dynamic MR-derived models are not yet in routineclinical use.

Faster and faster MRI techniques provide not only betterresolution in space and time, but also more and more dataavailable to the clinicians. Efforts should be made to providephysicians with only a small number of relevant parametersnecessary to establish a reliable diagnosis or fast staging of agiven disease. For cardiac motion analysis, the exact clinicalimportance of such detailed motion analysis has not beenfully answered, requiring extensive long-term studies in thisarea.

ACKNOWLEDGMENT

The authors would like to thank A. Aletras, J. Declerck,O. Faris, N. Osman, J. Prince, S. Sampath, M. Stone, R.Thompson, D. Unay, H. Wen, and B. Wyman for their con-tributions.

REFERENCES

[1] 2001 Heart and Stroke Statistical Update, Amer. Heart Assoc.(AHA), Dallas, TX, 2000.

[2] R. Muthupillai, D. J. Lomas, P. J. Rossman, J. F. Greenleaf, A. Man-duca, and R. L. Ehman, “Magnetic resonance elastography by directvisualization of propagating acoustic strain waves,”Science, vol.269, no. 5232, pp. 1854–1857, 1995.

[3] P. J. Basser and D. K. Jones, “Diffusion-tensor MRI: Theory,experimental design and data analysis—A technical review,”NMRBiomed, vol. 15, no. 7–8, pp. 456–467, 2002.

[4] D. Stark and W. G. Bradley,Magnetic Resonance Imaging, 3rded. St. Louis, MO: Mosby, Inc., 1999.

[5] E. M. Haacke, R. W. Brown, M. R. Thompson, and R. Venkatesan,Magnetic Resonance Imaging: Physical Principles and SequenceDesign. New York: Wiley, 1999.

[6] A. Amini and J. Prince,Measurement of Cardiac Deformations fromMRI: Physical and Mathematical Models. Norwell, MA: Kluwer,2001.

[7] P. C. Lauterbur, “Image formation by induced local interactions: Ex-amples employing nuclear magnetic resonance,”Nature, vol. 242, p.190, 1973.

[8] E. L. Hahn, “Spin echoes,”Phys. Rev., vol. 80, no. 4, p. 580, 1950.[9] A. N. Garroway, P. K. Grannel, and P. Mansfield, “Image formation

in NMR by selective irradiation,”J. Phys. C., vol. 7, p. L457, 1974.[10] D. B. Twieg, “Thek-trajectory formatulation of the NMR imaging

process with application in analysis and synthesis of imagingmethods,”Med. Phys, vol. 10, no. 5, p. 610, 1983.

[11] P. G. Gatehouse and D. N. Firmin, “The cardiovascular magnetic res-onance machine: Hardware and software requirements,”Herz, vol.25, no. 4, pp. 317–330, 2000.

[12] S. Ljunggren, “A simple graphical respresentation of Fourier-basedimaging methods,”J. Magn. Reson., vol. 54, p. 338, 1983.

[13] P. Mansfield, “Multi-planar image formation using NMR spinechoes,”J. Phys. C., vol. 10, p. L55, 1977.

[14] F. H. Epstein, S. D. Wolff, and A. E. Arai, “Segmentedk-space fastcardiac imaging using an echo-train readout,”Magn. Reson. Med.,vol. 41, pp. 609–613, 1999.

[15] C. B. Ahn, J. H. Kim, and Z. H. Cho, “High-speed spiral-scan echoplanar NMR imaging-I,”IEEE Trans. Med. Imag., vol. 5, pp. 2–7,Mar. 1986.

[16] C. H. Meyer, B. S. Hu, D. G. Nishimura, and A. Makovski, “Fastspiral coronary artery imaging,”Magn. Reson. Med., vol. 28, no. 2,pp. 202–213, 1992.

[17] D. K. Sodickson and W. Manning, “Simultaneous acquisition of spa-tial harmonics (SMASH): Fast imaging with radiofrequency coil ar-rays,”Magn. Reson. Med., vol. 38, pp. 591–603, 1997.

[18] K. P. Pruessmann, M. Weiger, M. B. Scheidegger, and P. Boesiger,“SENSE: Sensitivity encoding for fast MRI,”Magn. Reson. Med.,vol. 42, pp. 952–962, 1999.

[19] J. A. Feinstein, F. H. Epstein, A. E. Arai, T. K. F. Foo, R. S. Bal-aban, and S. D. Wolff, “Using cardiac phase to order reconstruc-tion (CAPTOR): A method to improve diastolic images,”J. Magn.Reson. Imag., vol. 7, no. 5, pp. 794–798, 1997.

[20] R. I. Pettigrew, J. N. Oshinski, G. Chatzimavroudis, and W. T.Dixon, “MRI techniques for cardiovascular imaging,”J. Magn.Reson. Imag., vol. 10, no. 5, pp. 590–601, 1999.

[21] S. B. Reeder and A. Z. Faranesh, “Ultrafast pulse sequence tech-niques for cardiac magnetic resonance imaging,”Top. Magn. Reson.Imag., vol. 11, no. 6, pp. 312–330, 2000.

[22] A. Lundberg, “Kinematics of the ankle and the foot:In vivostereophotogrammetry,”Acta Orthop. Scand., vol. 60, pp. 1–24,1989.

[23] J. J. Crisco, R. D. McGovern, and S. W. Wolfe, “Noninvasivetechnique for measuringin vivo three-dimensional carpal bonekinematics,”J. Ortho. Res., vol. 17, pp. 96–100, 1999.

[24] E. Stindel, J. K. Udupa, B. E. Hirsch, and D. Odhner, “A charac-terization of the geometric architecture of the peri-talar complex viaMRI: An aid to classification of feet,”IEEE Trans. Med. Imag., vol.18, pp. 753–763, Sept. 1999.

[25] , “An in vivo analysis of the peri-talar joint complex based onMR imaging,”IEEE Trans. Biomed. Eng., vol. 48, pp. 236–247, Feb.2001.

[26] R. C. Rhoad, J. J. Klimkiewicz, G. R. Williams, S. B. Kesmodel, J. K.Udupa, J. B. Kneeland, and J. P. Iannotti, “A newin vivo techniquefor 3D shoulder kinematics analysis,”Skeletal Radiol., vol. 27, pp.92–97, 1998.

[27] E. J. Van Langelaan, “A kinematic analysis of the tarsal joints: AnX-ray photogrammetric study,”Acta Ortho. Scand. (suppl.), vol.204, pp. 1–269, 1983.

[28] O. C. Morse and J. R. Singer, “Blood velocity measurement in intactsubjects,”Science, vol. 170, pp. 440–441, 1970.

[29] E. Zerhouni, D. Parish, W. Rogers, A. Yang, and E. Shapiro, “Humanheart: Tagging with MR imaging—a method for noninvasive assess-ment of myocardial motion,”Radiology, vol. 169, pp. 59–63, 1988.


[30] L. Axel and L. Dougherty, “MR imaging of motion with spatial mod-ulation of magnetization,”Radiology, vol. 171, pp. 841–845, 1989.

[31] T. J. Mosher and M. B. Smith, “A DANTE tagging sequence for theevaluation of translational sample motion,”Magn. Reson. Med, vol.15, no. 2, pp. 334–339, 1990.

[32] E. R. McVeigh and B. D. Bolster Jr., “Improved sampling of myocar-dial motion with variable seperation tagging,”Magn. Reson. Med.,vol. 39, pp. 657–661, 1998.

[33] E. R. McVeigh, “MRI of myocardial function: Motion tracking tech-niques,”Magn. Reson. Imag., vol. 14, pp. 137–150, 1996.

[34] W. S. Kerwin and J. L. Prince, “Ak-space analysis of MR tagging,”J. Magn. Reson., vol. 142, no. 2, pp. 313–322, 2000.

[35] A. J. de Crespigny, T. A. Carpenter, and L. D. Hall, “Cardiac taggingin the rat using a DANTE sequence,”Magn. Reson. Med., vol. 21,pp. 151–156, 1991.

[36] M. Stuber, S. E. Fischer, M. B. Scheidegger, and P. Boesiger, “To-ward high resolution myocardiac tagging,”Magn. Reson. Med., vol.41, pp. 639–643, 1999.

[37] S. E. Fischer, G. C. McKinnon, S. E. Maier, and P. Boesiger, “Im-proved myocardiac tagging contrast,”Magn. Reson. Med., vol. 30,pp. 191–200, 1993.

[38] P. A. Bottomley, T. H. Foster, R. E. Argersinger, and L. M. Pfeifer,“A review of normal tissue hydrogen NMR relaxation times andrelaxation mechanisms from 1–100 MHz: Dependence on tissuetype, NMR frequency, temperature, species, excision, and age,”Med. Phys., vol. 11, no. 4, pp. 425–448, 1984.

[39] G. Shechter, C. Ozturk, and E. R. McVeigh, “Interactive four-di-mensional segmentation of multiple image sets,” inProc. SPIE, vol.3976, Medical Imaging 2000: Image Display and Visualization, S.K. Mun, Ed., 2000, pp. 165–173.

[40] M. A. Guttman, J. L. Prince, and E. R. McVeigh, “Tag and contourdetection in tagged MR images of the left ventricle,”IEEE Trans.Med. Imag., vol. 13, pp. 74–88, Mar. 1994.

[41] M. A. Guttman, E. A. Zerhouni, and E. R. McVeigh, “Analysis ofcardiac function from MR images,”IEEE Comput. Graph. Appl.,vol. 17, pp. 30–38, Jan.–Feb. 1997.

[42] A. A. Young and L. Axel, “Three-dimensional motion and deforma-tion of the heart wall: Estimation with spatial modulation of magne-tization—a model based approach,”Radiology, vol. 185, no. 1, pp.241–247, 1992.

[43] A. A. Young, D. L. Kraitchman, L. Dougherty, and L. Axel,“Tracking and finite element analysis of stripe deformation inmagnetic resonance tagging,”IEEE Trans. Med. Imag., vol. 14, pp.413–421, Sept. 1995.

[44] J. Declerck, T. S. Denney, C. Ozturk, W. O‘Dell, and E. R. McVeigh,“Left ventricular motion reconstruction from planar tagged MRimages: A Comparison,”Phys. Med. Biol., vol. 45, no. 6, pp.1611–1632, 2000.

[45] W. G. O’Dell, C. C. Moore, W. C. Hunter, E. A. Zerhouni, and E. R.McVeigh, “Displacement field fitting for calculating 3D myocardialdeformations from planar tagged MR images,”Radiology, vol. 195,pp. 829–835, 1995.

[46] J. Declerck, N. Ayache, and E. R. McVeigh, “Use of a 4D plani-spheric transformation for the tracking and the analysis of LV motionwith tagged MR images,” presented at the SPIE Medical ImagingConf. 99, San Diego, CA, 1999.

[47] T. O’Donnell, T. Boult, and A. Gupta, “Global models with para-metric offsets as applied to cardiac motion recovery,” inProc. IEEEComput. Soc. Conf. Computer Vision Pattern Recognition, 1996, pp.293–299.

[48] J. Park, D. Metaxas, and L. Axel, “Analysis of left ventricular motionbased on volumetric deformable models and MRI-SPAMM,”Med.Image Anal., vol. 1, pp. 53–71, 1996.

[49] A. A. Young, D. Kraitchman, L. Dougherty, and L. Axel, “Trackingand finite element analysis of stripe deformation in magnetic res-onance imaging,”IEEE Trans. Med. Imag., vol. 14, pp. 413–421,Sept. 1995.

[50] T. S. Denney and E. R. McVeigh, “Model-free reconstructionof three-dimensional myocardial strain from planar tagged MRimages,”JMRI, vol. 7, no. 5, pp. 799–810, 1997.

[51] T. S. Denney, B. L. Gerber, and L. Yan, “Unsupervised reconstruc-tion of a three-dimensional left ventricular strain from paralleltagged cardiac images,”Magn. Reson. Med., vol. 49, no. 4, pp.743–754, 2003.

[52] M. Daniel, “Data fitting with B-spline curves,” inModeling andGraphics in Science and Technology, J. C. Teixeira and J. Rix,Eds. Berlin, Germany: Springer-Verlag, 1996.

[53] M. J. Moulton, L. L. Creswell, S. W. Downing, R. L. Actis, B. A.Szabo, M. W. Vannier, and M. K. Pasque, “Spline surface interpola-tion for calculating three-dimensional ventricular strains from MRItissue tagging,”Amer. J. Physiol., vol. 270, pp. H281–H297, 1996.

[54] A. A. Amini, Y. Chen, R. W. Curwen, V. Mani, and J. Sun, “CoupledB-snake grids and constrained thin-plate splines for analysis of 2-Dtissue deformations from tagged MRI,”IEEE Trans. Med. Imag., vol.17, pp. 344–356, June 1998.

[55] P. Radeva, A. A. Amini, and J. Huang, “Deformable B-solids andimplicit snakes for 3D localization and tracking of SPAMM MRIdata,”Comput. Vis. Image Understand., vol. 66, no. 2, pp. 163–178,1997.

[56] C. Ozturk and E. R. McVeigh, “Four dimensional B-spline basedmotion analysis of tagged cardiac MR images,” inProc. SPIE, vol.3660, Medical Imaging 1999: Physiology and Function from Multi-dimensional Images, C.-T. Chen and A. V. Clough, Eds., 1999, pp.46–56.

[57] , “Four dimensional B-spline based motion analysis of taggedMR images: Introduction andin vivo validation,”Phys. Med. Biol.,vol. 45, no. 6, pp. 1683–1702, 2000.

[58] , “Motion analysis of the whole heart,” inMeasurement of Car-diac Deformations from MRI: Physical and Mathematical Models,A. Amini and J. Prince, Eds. Norwell, MA: Kluwer, 2001, pp.91–117.

[59] T. Huang and A. A. Amini, “Volumetric motion tracking of left ven-tricle of the heart from tagged MRI by a B-spline solid model,” inProc SPIE, vol. 3660, Medical Imaging 1999: Physiology and Func-tion from Multidimensional Images, 1999, pp. 57–68.

[60] A. A. Young, “Model tags: Direct 3D tracking of heart motionfrom tagged MR images,” inLecture Notes in Computer Science,Medical Image Computing and Computer-Assisted Interven-tion—MICCAI. Heidelberg, Germany: Springer-Verlag, 1998,vol. 1496, pp. 92–101.

[61] N. F. Osman, W. S. Kerwin, E. R. McVeigh, and J. L. Prince, “Car-diac motion analysis using CINE harmonic phase (HARP) mag-netic resonance imaging,”Magn. Reson. Med., vol. 42, no. 6, pp.1048–1060, 1999.

[62] N. F. Osman and J. L. Prince, “Visualizing myocardial function usingHARP MRI,” Phys. Med. Biol., vol. 45, no. 6, pp. 1665–1682, 2000.

[63] N. F. Osman, E. R. McVeigh, and J. L. Prince, “Imaging heart motionusing harmonic phase MRI,”IEEE Trans. Med. Imag., vol. 19, pp.186–201, Mar. 2000.

[64] N. F. Osman and J. L. Prince, “Harmonic phase MRI,” inMeasure-ment of Cardiac Deformations from MRI: Physical and Mathemat-ical Models, A. Amini and J. Prince, Eds. Norwell, MA: Kluwer,2001, pp. 119–150.

[65] S. Isci and C. Ozturk, “Automatic myocardial localization for taggedMRI,” in Proc. 10th Int. Soc. Magnetic Resonance Medicine, 2002,p. 2440.

[66] D. J. Bryant, J. A. Payne, D. N. Firmin, and D. B. Longmore, “Mea-surement of flow with NMR imaging using a gradient pulse andphase difference technique,”J. Comput. Assist. Tomog., vol. 8, pp.588–593, 1984.

[67] G. L. Nayler, D. N. Firmin, and D. B. Longmore, “Blood flowimaging by cine magnetic resonance,”J. Comput. Assist. Tomog.,vol. 10, pp. 715–722, 1986.

[68] V. J. Wedeen, “Magnetic resonance imaging of myocardial kine-matics, technique to detect, localize, and quantify the strain ratesof the active human myocardium,”Magn. Reson. Med., vol. 27, pp.52–67, 1992.

[69] Y. Zhu, M. Drangova, and N. J. Pelc, “Estimation of deformationgradient from cine-PC velocity data,”IEEE Trans. Med. Imag., vol.16, pp. 840–851, Dec. 1997.

[70] , “Fourier tracking of myocardial motion using cine-PC data,”Magn. Reson. Med., vol. 35, pp. 471–480, 1996.

[71] Y. Zhu and N. J. Pelc, “A spatiotemporal model of cyclic kinematicsand its application to analysing nonrigid motion with MR velocityimages,”IEEE Trans. Med. Imag., vol. 18, pp. 557–569, July 1999.

[72] , “Three-dimensional motion tracking with volumetric phasecontrast CMR velocity imaging,”J. Magn. Reson. Imag., vol. 9, pp.111–118, 1999.

[73] A. E. Arai, C. C. Gaither 3rd, F. H. Epstein, R. S. Balaban, andS. D. Wolff, “Myocardial velocity gradient imaging by phase con-trast CMRI with application to regional function in myocardial is-chaemia,”Magn. Reson. Med., vol. 42, pp. 98–109, 1999.

[74] G. Suryan, “Nuclear resonance in flowing liquids,”Proc. IndianAcad. Sci., vol. 33, p. 107, 1951.


[75] J. R. Singer, “Blood flow rates by nuclear magnetic resonance mea-surements,”Science, vol. 130, pp. 1652–1653, 1959.

[76] E. L. Hahn, “The detection of sea water motion by nuclear preces-sion,” J. Geophys. Res., vol. 65, no. 2, pp. 776–777, 1960.

[77] E. O. Stejskal, “Use of spin echoes in a pulsed magnetic field gra-dient to study anisotropic, restricted diffusion and flow,”J. Chem.Phys., vol. 43, no. 10, pp. 3597–3603, 1965.

[78] T. Grover and J. R. Singer, “NMR spin-echo flow measurements,”J. Appl. Phys., vol. 42, pp. 938–940, 1971.

[79] A. Caprihan, R. H. Griffey, and E. Fukushima, “Velocity imaging ofslow coherent flows using stimulated echoes,”Magn. Reson. Med.,vol. 15, pp. 327–333, 1990.

[80] A. H. Aletras, S. Ding, R. S. Balaban, and H. Wen, “DENSE: Dis-placement encoding with stimulated echoes in cardiac functionalMRI,” J. Magn. Reson., vol. 137, pp. 247–252, 1999.

[81] S. Sampath, J. A. Derbyshire, N. F. Osman, E. Atalar, and J. L.Prince, “Real-time imaging of cardiac strain using an ultra-fastHARP sequence,” inProc. 9th Int. Soc. Magnetic ResonanceMedicine, 2001, p. 111.

[82] A. H. Aletras, R. S. Balaban, and H. Wen, “High-resolution strainanalysis of the human heart with fast-DENSE,”J. Magn. Reson.,vol. 140, pp. 41–57, 1999.

[83] A. H. Aletras and H. Wen, “Mixed echo train acquisition displace-ment encoding with stimulated echoes: An optimized DENSEmethod forin vivo functional imaging of the heart,”Magn. Reson.Med., vol. 46, pp. 523–534, 2001.

[84] J. Frahm, K. D. Merboldt, W. Hanicke, and A. Haase, “Stimulatedecho imaging,”J. Magn. Reson., vol. 64, pp. 81–93, 1985.

[85] J. Frahm, W. Hanicke, H. Bruhn, M. L. Gyngell, and K. D. Merboldt,“High speed STRAM-MRI of the human heart,”Magn. Reson. Med.,vol. 22, pp. 133–142, 1991.

[86] S. Sampath, J. A. Derbyshire, N. F. Osman, and J. L. Prince,“Real-time imaging of myocardial strain patterns using a fastHARPsequence with CSPAMM,” inProc. Int. Soc. Magnetic ResonanceMedicine, 2002, p. 1631.

[87] E. G. Walsh, M. Doyle, M. E. Kortright, I. M. Straeter-Knowlen,and G. M. Pohost, “Recent progress in radiofrequency-tagged leftventricular function studies,”J. Cardiovasc. Magn. Reson., vol. 1,no. 2, pp. 185–193, 1999.

[88] C. C. Moore, E. R. McVeigh, and E. A. Zerhouni, “Quantitativetagged magnetic resonance imaging of the normal human left ven-tricle,” Top. Magn. Reson. Imag., vol. 11, no. 6, pp. 359–371, 2000.

[89] C. C. Moore, C. H. Lugo-Olivieri, E. R. McVeigh, and E. A.Zerhouni, “Three-dimensional systolic strain patterns in the normalhuman left ventricle: Characterization with tagged MR imaging,”Radiology, vol. 214, no. 2, pp. 453–466, 2000.

[90] C. C. Moore, E. R. McVeigh, and E. A. Zerhouni, “Noninvasive mea-surement of three-dimensional myocardial deformation with taggedmagnetic resonance imaging during graded local ischemia,”J. Car-diovasc. Magn. Reson., vol. 1, no. 3, pp. 207–222, 1999.

[91] M. Stuber, M. B. Scheidegger, S. E. Fischer, E. Nagel, F. Steine-mann, O. M. Hess, and P. Boesiger, “Alterations in the local myocar-dial motion pattern in patients suffering from pressure overload dueto aortic stenosis,”Circulation, vol. 100, no. 4, pp. 361–368, 1999.

[92] A. S. Blom, J. J. Pilla, S. V. Pusca, H. J. Patel, L. Dougherty, Q. Yuan,V. A. Ferrari, L. Axel, and M. A. Acker, “Dynamic cardiomyoplastydecreases myocardial workload as assessed by tissue tagged MRI,”ASAIO J., vol. 46, no. 5, pp. 556–562, 2000.

[93] R. Mankad, C. J. McCreery, W. J. Rogers Jr, R. J. Weichmann, E. B.Savage, N. Reichek, and C. M. Kramer, “Regional myocardial strainbefore and after mitral valve repair for severe mitral regurgitation,”J. Cardiovasc. Magn. Reson., vol. 3, no. 3, pp. 257–266, 2001.

[94] P. Dubach, J. Myers, P. Bonetti, T. Schertler, V. Froelicher, D.Wagner, M. Scheidegger, M. Stuber, R. Luchinger, J. Schwitter,and O. Hess, “Effects of bisoprolol fumarate on left ventricularsize, function, and exercise capacity in patients with heart failure:Analysis with magnetic resonance myocardial tagging,”Amer.Heart J., vol. 143, no. 4, pp. 676–683, 2002.

[95] B. T. Wyman, W. C. Hunter, F. W. Prinzen, and E. R. McVeigh,“Mapping propagation of mechanical activation in the paced heartwith MRI tagging,” Amer. J. Physiol., vol. 276, pp. H881–H891,1999.

[96] D. A. Kass, C. H. Chen, C. Curry, M. Talbot, R. Berger, B. Fetics,and E. Nevo, “Improved left ventricular mechanics from acute VDDpacing in patients with dilated cardiomyopathy and ventricular con-duction delay,”Circulation, vol. 99, pp. 1567–1573, 1999.

[97] F. W. Prinzen, W. C. Hunter, B. T. Wyman, and E. R. McVeigh,“Mapping of regional myocardial strain and work during ventric-ular pacing: Experimental study using magnetic resonance imagingtagging,”J. Amer. Coll. Cardiol., vol. 33, pp. 1735–1742, 1999.

[98] G. L. Freeman, M. M. LeWinter, R. L. Engler, and J. W. Covell, “Re-lationship between myocardial fiber direction and segmental short-ening in the midwall of the canine left ventricle,”Circulation, vol.56, pp. 31–39, 1985.

[99] R. B. Thompson and E. R. McVeigh, “High temporal resolutionmulti-echo phase contrast MRI,”Magn. Reson. Med., vol. 47, no.3, pp. 499–512, 2002.

[100] M. Stone, “Laboratory techniques for investigating speech articula-tion,” in The Handbook of Phonetic Sciences, W. J. Hardcastle andJ. Laver, Eds. New York: Blackwell, 1997, ch. 1.

[101] S. Kiritani, K. Ito, and O. Fujimura, “Tongue-pellet tracking by acomputer-controlled X-ray micro beam system,”J. Acoust. Soc.Amer., pt. 2, vol. 57, no. 6, pp. 1516–1520, 1975.

[102] J. M. Palmer, “Dynamic palatography: General implications of locusand sequencing patterns,”Phonetica, vol. 28, pp. 76–85, 1973.

[103] M. Stone, A. Faber, and M. Cordaro, “Cross-sectional tongue move-ment and tongue palate movement patterns in [s] and [sh] sylla-bles,” in Proc. 12th Int. Congress Phonetic Sciences, vol. 2, 1991,pp. 354–357.

[104] P. MacNeiladge and G. Sholes, “An electromyography study of thetongue during vowel production,”J. Speech Hearing Res., vol. 7, pp.209–232, 1964.

[105] K. Honda and N. Kusakawa, “Compatibility between auditory andrepresentations of vowels,”Act. Otolaryngol. (Stockholm) Suppl.,vol. 532, pp. 103–105, 1997.

[106] Y. S. Akgul, C. Kambhamettu, and M. Stone, “Automatic extractionand tracking of the tongue contours,”IEEE Trans. Med. Imag., vol.18, pp. 1035–1045, Oct. 1999.

[107] W. M. Kier and K. K. Smith, “Tongues, tentacles, and trunks: Thebiomechanics of movement in muscular-hydrostats,”Zoo. J. Linn.Soc., vol. 83, pp. 307–324, 1985.

[108] T. Baer, J. C. Gore, S. Boyce, and P. W. Nye, “Application of MRIto the analysis of speech production,”Magn. Reson. Imag., vol. 5,pp. 1–7, 1987.

[109] D. Demolin, T. Metens, and A. Soquet, “Real-time MRI and ar-ticulator coordinations in vowels,” inProc. ICSLP 96, 1998, pp.272–275.

[110] D. Dick, C. Ozturk, A. Douglas, E. R. McVeigh, and M. Stone,“Three-dimensional tracking of tongue motion using tagged MRI,”in Proc. 8th Int. Soc. Magnetic Resonance Medicine, 2000, p. 553.

[111] C. Ozturk, M. Stone, D. Unay, A. Lundberg, and E. R. McVeigh,“Real-time MR imaging with tagging for tongue motion analysisduring speech,” inProc. 9th Int. Soc. Magnetic Resonance Medicine,April 2001, p. 1609.

[112] W. A. Boxley and T. J. Reeves, “Abnormal regional myocardial per-formance in coronary artery disease,”Prog. Cardiovasc. Dis., vol.13, pp. 405–421, 1971.

[113] M. V. Herman and R. Gorlin, “Implications of left ventricular syn-ergy,” Amer. J. Cardiol., vol. 23, pp. 538–547, 1969.

[114] E. McVeigh, “Regional myocardial function,”Cardiac Magn.Reson. Imag., vol. 16, no. 2, pp. 189–206, 1998.

[115] M. I. Miyamoto, “Abnormal global left ventricular relaxation oc-curs early during the development of pharmacologically induced is-chemia,”J. Amer. Soc. Echocardiogr., vol. 12, no. 2, pp. 113–120,1999.

[116] J. Barkhausen, M. Goyen, S. G. Ruhm, H. Eggebrecht, J. F. De-batin, and M. E. Ladd, “Assessment of ventricular function withsingle breath-hold real-time steady-state free precession cine MRimaging,” AJR Amer. J. Roentgenol., vol. 178, no. 3, pp. 731–735,2002.

[117] A. Pentland and B. Horowitz, “Recovery of nonrigid motion andstructure,”IEEE Trans. Pattern Anal. Machine Intell., vol. 13, pp.730–742, July 1991.

[118] N. Ayache, I. Cohen, and I. Herlin,Active Vision, A. Blake andA. Yuille, Eds. Berlin, Germany: Springer-Verlag, 1992, pp.285–302.

[119] T. McInerney and D. Terzopoulos, “A dynamic finite element sur-face model for segmentation and tracking in multidimensional med-ical images with application to cardiac 4D image analysis,”Comput.Med. Imag. Graph., vol. 19, pp. 69–83, 1995.

[120] A. Amini, R. Owen, P. Anandan, and J. Duncan, “Non-rigidmotion models for tracking the left ventricular wall,” inLectureNotes in Computer Science, Information Processing in MedicalImages. Heidelberg, Germany: Springer-Verlag, 1991, vol. 511,pp. 343–357.


[121] J. Duncan, R. Owen, L. Staib, and E. Anandan, “Measurement ofnonrigid motion using contour shape descriptors,” inProc. IEEEComput. Soc. Conf. Computer Vision Pattern Recognition, 1991, pp.318–324.

[122] D. Friboulet, I. Magnin, C. Mathieu, A. Pommert, and K. Hoehne,“Assessment and visualization of the curvature of the left ventriclefrom 3-D medical images,”Comput. Med. Imag. Graph., vol. 17, pp.257–262, 1993.

[123] S. Benayoun, C. Nastar, and N. Ayache, “Dense nonrigid motion es-timation in sequences of 3D images using differential constraints,”in Lecture Notes in Computer Science, Computer Vision, Virtual Re-ality and Robotics in Medicine. Heidelberg, Germany: Springer-Verlag, 1995, vol. 905, pp. 309–318.

[124] E. Shi, A. Amini, G. Robinson, A. Sinusas, C. Constable, and J.Duncan, “Shape-based 4D left ventricular myocardial functionanalysis,” inIEEE Workshop Biomedical Image Analysis, 1994, pp.88–97.

[125] E. Meyer, R. Constable, A. Sinusas, and J. Duncan, “Trackingmyocardial deformation using spatially-constrained velocities,” inProc. 14th Int. Conf. Information Processing Medical Imaging,1995, pp. 177–188.

[126] J. McEachen, E. Meyer, R. Constable, A. Nehorai, and J. Duncan,“A recursive filter for phase velocity assisted shape-based tracking ofcardiac nonrigid motion,” inProc. IEEE Int. Conf. Computer Vision,1995, pp. 653–658.

[127] P. G. Danias, M. Stuber, N. A. Tritos, C. Salton, K. V. Kissinger, andW. J. Manning, “CSPAMM assessment of systolic and diastolic leftventricular apical rotation in obesity,” inProc. 8th Int. Soc. MagneticResonance Medicine, 2000, p. 1613.

[128] C. H. Lorenz, K. J. Lunn, C. Dent, and S. A. Wickline, “Ventriculartwist allows sensitive detection of early-stage diabetic cardiomy-opathy in the rat,” inProc. 8th Int. Soc. Magnetic ResonanceMedicine, 2000, p. 1606.

[129] J. M. Declerck, C. Ozturk, L. Gutirrez, G. Shechter, and E. R.McVeigh, “Combined analysis of ventricular and atrial motionusing MRI,” in Proc. 9th Int. Soc. Magnetic Resonance Medicine,2001, p. 1869.

[130] O. Faris, F. Evans, C. Ozturk, D. Ennis, J. Taylor, and E. R.McVeigh, “Correlation between electrical and mechanical activa-tion in the paced canine heart,” inProc. 9th Int. Soc. MagneticResonance Medicine, 2001, p. 1842.

[131] E. R. McVeigh and C. Ozturk, “Imaging myocardial strain,”IEEESignal Processing Mag., vol. 18, pp. 44–56, Nov. 2001.

[132] C. Ozturk and E. R. McVeigh, “Motion analysis of both ventriclesusing tagged-MRI in paced canine hearts,” inProc. 8th Int. Soc.Magnetic Resonance Medicine, 2000, p. 1612.

[133] , “Motion analysis of both ventricles using tagged-MRI,” inProc. SPIE, vol. 3978, Medical Imaging 2000: Physiology and Func-tion from Multidimensional Images, C. T. Chen and A. V. Clough,Eds., 2000, pp. 225–232.

[134] A. H. Aletras, G. S. Tilak, S. Ambati, D. Ennis, C. Ozturk, and A.E. Arai, “Displacement and strain measurements in the human heartwith SPAMM and meta-DENSE,” inProc. 10th Int. Soc. MagneticResonance Medicine, 2002, p. 78.

[135] M. Stone, D. Dick, E. Davis, A. Douglas, and C. Ozturk, “Modelingthe internal tongue using principal strains,” inProc. 5th Speech Pro-duction Seminar, 2000, pp. 133–136.

[136] H. K. Huszar, C. Ozturk, G. Schecter, H. R. Halperin, and A. C.Lardo, “Three-dimensional tracking of chest wall deformationduring cardiopulmonary resuscitation using tagged MRI,” inProc.9th Int. Soc. Magnetic Resonance Medicine, 2001, p. 188.

[137] V. J. Napadow, V. Mai, A. Bankier, R. J. Gilbert, R. Edelman, andQ. Chen, “Related articles determination of regional pulmonaryparenchymal strain during normal respiration using spin inversiontagged magnetization MRI,”J. Magn. Reson. Imag., vol. 13, no. 3,pp. 467–474, 2001.

[138] S. C. Wayte, T. W. Redpath, and D. J. Beale, “A cine-SPAMMsequence for NMR imaging of pulsatile cerebrospinal fluid flow,”Phys. Med. Biol., vol. 38, no. 3, pp. 455–463, 1993.

[139] S. C. Wayte and T. W. Redpath, “Cine magnetic resonance imagingof pulsatile cerebrospinal fluid flow using CSPAMM,”Br. J. Radiol.,vol. 67, no. 803, pp. 1088–1095, 1994.

[140] S. K. Lee, T. S. Chung, and Y. S. Kim, “Evaluation of CSFmotion in syringomyelia with spatial modulation of magnetization(SPAMM),” Yonsei Med. J., vol. 43, no. 1, pp. 37–42, 2002.

[141] M. Rentsc, “Dynamic magnetic resonance imaging defecography: Adiagnostic alternative in the assessment of pelvic floor disorders inproctology,”Dis. Colon Rectum, vol. 44, no. 7, pp. 999–1007, 2001.

[142] K. Singh, “Assessment and grading of pelvic organ prolapse by useof dynamic magnetic resonance imaging,”Amer. J. Obstet. Gynecol.,vol. 185, no. 1, pp. 71–77, 2001.

[143] C. McCabe, “Trauma: An annotated bibliography of the recent lit-erature 2000,”Amer. J. Emerg. Med., vol. 19, no. 5, pp. 437–452,2001.

[144] P. D. Franklin, “Accuracy of imaging the menisci on an in-office,dedicated, magnetic resonance imaging extremity system,”Amer. J.Sports Med., vol. 25, no. 3, pp. 382–388, 1997.

[145] A. H. McGregor, “Assessment of spinal kinematics using open in-terventional magnetic resonance imaging,”Clin. Orthop., vol. 392,pp. 341–348, 2001.

[146] F. T. Sheehan, “Quantitative MR measures of three-dimensionalpatellar kinematics as a research and diagnostic tool,”Med. Sci.Sports Exerc., vol. 31, no. 10, pp. 1399–1405, 1999.

[147] R. J. Miller, “Wrist MRI and carpal instability: What the surgeonneeds to know, and the case for dynamic imaging,”Semin. Muscu-loskelet. Radiol., vol. 5, no. 3, pp. 235–240, 2001.

[148] P. J. Keir, “Magnetic resonance imaging as a research tool for biome-chanical studies of the wrist,”Semin. Musculoskelet. Radiol., vol. 5,no. 3, pp. 241–250, 2001.

[149] A. S. Arnold, “Accuracy of muscle moment arms estimated fromMRI-based musculoskeletal models of the lower extremity,”Comput. Aided Surg., vol. 5, no. 2, pp. 108–119, 2000.

Cengizhan Ozturk (Member, IEEE) receivedthe M.D. degree from Marmara School ofMedicine, Istanbul, Turkey, in 1990, the special-ization in physiology from Cerrahpasa MedicalSchool, Istanbul University, Istanbul, Turkey,in 1994, and the Ph.D. degree from the Schoolof Biomedical Engineering, Science and HealthSystems, Drexel University, Philadelphia, PA,in 1997, with a dissertation on the developmentof a new 3-D computer vision technique (perfectsubmap based structured light) to acquire 3-D

surface images of objects and finding better methods of aligning them.From 1997 to 2000, he was a Postdoctoral Research Fellow at Medical

Imaging Lab, Department of Biomedical Engineering, Johns Hopkins Uni-versity, Baltimore, MD, where his main research area was the developmentof new methods for rapid physiologic evaluation and tissue tracking for car-diac MRI images, evaluation of right ventricular function, atrial motion, anddevelopment of a four-chamber motion field. Since 2000, he has been an As-sistant Professor at the Institute of Biomedical Engineering, Bogazici Uni-versity, Istanbul, Turkey. He is currently performing research on real-timeMRI and magnetic resonance guided medical interventions at the NationalInstitutes of Health, National Heart, Lung, and Blood Institute, Bethesda,MD.

J. Andrew Derbyshire received the B.S. degreein mathematics and the M.S. degree in controlengineering from the University of Sheffield,Sheffield, U.K., in 1988 and 1989, respectively,and the Ph.D. degree from University of Cam-bridge, Cambridge, U.K., in 1995. His Ph.D.research on echo-planar and motion sensitiveMRI methods was conducted at the HerchelSmith Laboratory, University of Cambridge,under Dr. T. A. Carpenter and Professor L. D.Hall.

From 1995 to 1998, he was a Postdoctoral Fellow with the Imaging Re-search Group, Department of Medical Biophysics, University of Toronto,Toronto, ON, Canada, working with Drs. M. Henkelman, S. Hinks, and G.Wright on the development of MRI technologies for surgical intervention, ina joint project with General Electric Medical Systems, Toronto, ON, Canada.From 1998 to 2000, he was an Advanced Development Engineer with Gen-eral Electric Medical Systems, developing cardiovascular MRI as a clinicalproduct while working as a visiting scientist in the Radiology Department,School of Medicine, Johns Hopkins University, Baltimore, MD. In 2000, hejoined Dr. E. McVeigh’s group as a Staff Scientist, Laboratory of CardiacEnergetics, National Institutes of Health, National Heart, Lung, and BloodInstitute, Bethesda, MD. His research interests include real-time and inter-active MR imaging technologies focusing on cardiovascular applications.


Elliot R. McVeigh (Member, IEEE) receivedthe B.Sc. degree in physics and the Ph.D. degreefrom the University of Toronto, Toronto, ON,Canada, in 1984 and 1988, respectively. Heobtained his Ph.D. under the guidance of Drs.Mark Henkelman and Michael Bronskill at theOntario Cancer Institute, Toronto, ON, Canada,where one of the firstin vivo MRI systems inCanada was installed.

From 1988 to 1999, he was with the radiologyfaculty at Johns Hopkins University School of

Medicine, Baltimore, MD, working with Dr. E. Zerhouni developing aresearch program in cardiac MRI. In 1991, he joined the Department ofBiomedical Engineering, Johns Hopkins, to found the Medical ImagingProgram. In 1999, he established a laboratory for the development ofcardiovascular interventional MRI at the National Institutes of Health,National Heart, Lung, and Blood Institute, Bethesda, MD. He has publishedover 100 peer-reviewed research papers.


Documents

Estimating Motion From MRI Dataee-classes.usc.edu/ee691/read/Ozturk2003p1268.pdfthe user a known pattern to detect for motion tracking. The MRI signal is a modulated complex quantity,