Upload
vuquynh
View
213
Download
0
Embed Size (px)
Citation preview
CONVERSION OF ECG GRAPH INTO DIGITAL FORMAT AND DETECTING THE DISEASE
G. Angelo Virgin1, Dr.M.Sangeetha2
1Assistant Professor, ECE, Bharath University, Chennai, TN, India
2Professor, Department Of Electronics and Communication Engineering,
Bharath University, Selaiyur, Chennai-73, India 1 [email protected],
Abstract:Electrocardiogram (ECG) is the most
important and widely used method to study the heart
related diseases. The detailed study of ECG graph by
the medical practitioner helps him to understand and
identify the condition of the heart. Based on the
information retrieved from the ECG graph the patient
can be given proper treatment. The person having a
medical history of heart ailments will have to
maintain a record of all the ECG papers for timely
analysis and diagnosis of the diseases. This process
requires large storage space and extensive manual
effort. The conventional technique of visual analysis
to inspect the ECG signals by doctors or physicians
are not effective and time consuming. Therefore, an
automatic system which involves digital signal
integration and analysis is required. We presents a
digital signal processing tool developed using
MATLAB, which provides a very low-cost and
effective strategy for Analog to Digital conversion.
This digitization process is generated using gradient
based feature extraction algorithm without using a
dedicated hardware. In addition, this tool can be used
to potentially integrate digitized ECG information
with digital ECG analysis programs and with the
patient's disease analysis.
Keywords: ECG, Compression, Segmentation, image
retrieval, Digitization, Laplacian filtering.
1. Introduction
Digital equipment are flexibility of working their
output. Electro cardiogram signals recorded to digital
format. Therefore digital signals are high signal
processing retrieval for the information. The
advantages of ECG signal are digital technology for
turns to first- choice. Therefore digital technologies
are high cost for modern digital equipment. The
serious barrier are crossed by working analogical
device limitations. The analog devices are used to
convert the analog values for suitable interface to
digital microcomputer. The digitizing analogical data
are many areas, medicine history of patients would be
kept and case studies are correlated. The storage ECG
signals are inefficient data are processing for
unfeasible. The designing acquisition values are
interface to microcomputers and instead of purchasing
sophisticated computerized equipment. The trivial task
are assemble to design for ECG signals. The MATLAB
software tool is used analog version of signals and
spectra digitalized ECG scanning data file are efficiently
processed and stored. Therefore alternative approach for
A/D conversion without requiring specific hardware.
2. ECG Pattern Recognition
ECG processing proposed to pattern recognition,
parameter extraction, spectro- temporal techniques for
assessment for the heart’s status [7], de noising, baseline
correction and arrhythmia detection. Three basic waves
are P, QRS and T (Fig 1). These waves are far field
induced by specific electrical phenomena cardiac
surface, namely atrial depolarization (P), ventricular
depolarization (QRS) and ventricular depolarization (T).
Numerous techniques are developed to recognize and
analyse waves from digital filtering techniques are neural
network (NN) and spectro-temporal techniques. A great
deal of research devoted to digital processing of ECG
waves, are QRS, P and T wave detection (Fig 1). The P
wave detection is the problem (especially in the presence
of noise in the case of ventricular tachycardias).
QRS and T wave detection have been developed.
In this cases clinically acceptable results arrhythmia
analysis are received a great deal of attention, directed to
QRS and PVC classification and supra ventricular
arrhythmias. In shifted from arrhythmia to ischemia
detection and monitoring.
It causes depolarization for resting membrane
potential of ischemic region with respect to resting
membranee potential of normal region. The potential
difference between the flows of injury current is
manifested in the ECG by an elevated or depressed ST
segment (Fig 1) [10-14]. ST segment detection and
characterization of major problem of ECG processing
inhibiting factors are slow baseline drift, noise, sloped
ST, patient dependent abnormal ST depression levels,
and varying ST-T patterns in the ECG patient. A number
of methods proposed in the literature for ST detection
based on digital filtering, analysis of first derivative
signal, and syntactic methods. To measure specific
parameters in ways critically dependent upon the correct
International Journal of Pure and Applied MathematicsVolume 116 No. 15 2017, 465-471ISSN: 1311-8080 (printed version); ISSN: 1314-3395 (on-line version)url: http://www.ijpam.euSpecial Issue ijpam.eu
465
detection of J point on the ECG, which is the
inflection point following the S wave. Where ST
segment is sloped, or the signal is noisy, it is
impossible to identify this point with reliability (Fig
1).
Figure 1. Normal and abnormal ECG Waveform.
The constituent ECG waves and J point. In the
ischemic ECG, we observe the ST elevation, and
observe the second beat that the J point is not easily
discernible. The biomedical signal processing are
based on standard in the normal case, annotated
databases and forming a common reference. The
database are ischemia detection in the MIT
database can be used. The performance of ischemic
beats episodes performance are often used [11
3. Proposed Work
The electrocardiogram is a signal functional
investigation. The electric signals are generated
human heart and create cardiac cycle generate blood
circulation. The basic components are P wave, QRS
complex and T wave as show in (fig 2) P wave is
generated atrium depolarization. The QRS complex is
generated for ventricle depolarization and T wave
generated from cardiovascular disease is one of the
leading causes of death. The lacking of physician
expert for analysis on ECG signal is serious problem
especially on rural area. Therefore ECG classification
are ECG printout is relieve this problem. The example
of paper based ECG figure.1
detection of J point on the ECG, which is the
ection point following the S wave. Where ST
segment is sloped, or the signal is noisy, it is
impossible to identify this point with reliability (Fig
Normal and abnormal ECG Waveform.
The constituent ECG waves and J point. In the
ECG, we observe the ST elevation, and
observe the second beat that the J point is not easily
discernible. The biomedical signal processing are
based on standard in the normal case, annotated
databases and forming a common reference. The
ia detection in the MIT-BIH
database can be used. The performance of ischemic
beats episodes performance are often used [11-15].
The electrocardiogram is a signal functional
investigation. The electric signals are generated from
human heart and create cardiac cycle generate blood
circulation. The basic components are P wave, QRS
complex and T wave as show in (fig 2) P wave is
generated atrium depolarization. The QRS complex is
generated for ventricle depolarization and T wave is
generated from cardiovascular disease is one of the
leading causes of death. The lacking of physician
expert for analysis on ECG signal is serious problem
especially on rural area. Therefore ECG classification
The example
Figure 2. Example image of paper based ECG printout.
In this paper proposed filters used to reduce
influence of noise, such as muscle noise, power line and
baseline wonder. New database are composed with each
type of beat annotations such as -
Left bundle branch block beat and 1 for Right bundle
beat and 2 for premature ventricular contraction [16
Figure 3. Generic ECG digitization process
Scanning
ECG paper recordings are scanned. The sca
resolution are 600/400/200 dots per inch. Preferred
algorithm is compression for the JPEG image. Figure1
represents scanned ECG paper recording.
Crop region of interest
ECG graph is selected the particular region of
interest to obtain the data. In order to retrieved the
information from the ECG image[23
Image Binarization
It is a process and translate for colour image into
a binary image. It is widespread process for image
processing and especially images are contain neither
artistic nor iconographic. In binarization operation
reduces to number of colours to a binary level are
apparent gains of storage and besides simplifying the
image analysis as compared to the true colour image
processing. The axis are composing the image and
applied binarization process by Otsu’s algorithm, since it
has been shown to provide satisfactory results in many
applications. Binarization is an important step for
moving from a biomedical map image towards a digital
signal are target of the process obtain a sequence of
values that corresponded amplitude of a uniform time
series.
Gradient based Feature Extraction
The gradient is a vector which has both
magnitude and directions.
Example image of paper based ECG printout.
In this paper proposed filters used to reduce
influence of noise, such as muscle noise, power line and
baseline wonder. New database are composed with each
-1 for Normal, -2 for
Left bundle branch block beat and 1 for Right bundle
beat and 2 for premature ventricular contraction [16-22].
Generic ECG digitization process
ECG paper recordings are scanned. The scanning
resolution are 600/400/200 dots per inch. Preferred
algorithm is compression for the JPEG image. Figure1
represents scanned ECG paper recording.
ECG graph is selected the particular region of
n the data. In order to retrieved the
information from the ECG image[23-30].
It is a process and translate for colour image into
a binary image. It is widespread process for image
processing and especially images are contain neither
tistic nor iconographic. In binarization operation
reduces to number of colours to a binary level are
apparent gains of storage and besides simplifying the
image analysis as compared to the true colour image
processing. The axis are composing the image and then
applied binarization process by Otsu’s algorithm, since it
has been shown to provide satisfactory results in many
applications. Binarization is an important step for
moving from a biomedical map image towards a digital
s obtain a sequence of
values that corresponded amplitude of a uniform time
Gradient based Feature Extraction
The gradient is a vector which has both
International Journal of Pure and Applied Mathematics Special Issue
466
De-Skewing
It is a common phenomenon can be appear
scanned image. Skewing image to an angle and
results are rotated image. Hough Transform used to
de-skew image. Skew angle is calculated using
background grid lines of scanned ECG image.
Image Enhancement
In this step enhances to ECG image by
making the signal lines. The filtering is applied to
making background noise lighter than the ECG signal.
Threshold values are chosen by comparing between
noise pixels representing actual ECG signal. ECG
signal pixel values are closer threshold and pixels are
made by subtracting fixed value. The noise pixels are
close to the threshold, will be made lighter by adding
a fix value. Therefore image will contain distinct
ECG signal in the image.
Colour Based Segmentation
To remove background grid of an ECG
lighter shade of colour actual signal waveform are
processed column by column. The darkest pixels are
extracted each column and row are replaced black
pixel and rest of the pixels as pure white. However
lead to extraction undesired printed character as well.
Fig 4 shows output of colour based segmentation; it
has some printed characters also.
Region Based Segmentation
Removes the isolated pixels that do not
represent the signal. The following steps are
Step 1: Remove the frame of image is border of the
input image.
Step 2: Scan the input ECG image is column by
column.
Step 3: Repeat each column:
a. Extract each black pixels
b. In case of isolated pixels delete them.
c. If more than one black pixel:
i. Algorithm checks previous and next columns. They
contain one pixel, then the current column select pixel
that likely goes flow of signal pattern.
ii. The next column contains the same problem of
having black pixel for the current column select the
middle pixel that lies exactly in the middle of the
black pixel pattern.
Step 4: Save index of the picked pixel. The process of
choosing middle value in step 2 is performed by
following algorithm:
Step 4.1: The centres of different regions
Step 4.2: separate regions and compute centroid of each
region.
Figure 4. ECG Steps
1a: The threshold value region are separation. If
current_point>previous_sample + threshold_value add
centre.
Step 5: Pick centre point value of closest to previous
column pixel index. In this centres calculated in step 2
are used. To find the centre that has minimum difference
between previous column pixel and that next centre.
Signal Representation
The ECG signals are corresponding column with
pixel. Finally the signals are saved as a txt file.
Median Filtration
Actual ECG signal recorded an ECG machine
can be degraded by presence of noise. Various sources of
noise are Power Line Interface (50Hz or 60Hz noise
from power lines) Baseline wander.
Muscle noise
Other interference noise signal are misdiagnosis
desirable to remove noise and the actual ECG signal. A
noisy ECG signals are taken the performance of various
filters including median filter. FIR filters are compared
to found that median filter has best result both by visual
inspection and quantitative measurement.
Axis Identification
It is different types of data plotting are analysed
for the sake of generality of the methodology proposed.
This paper used to register the data contain grid and box.
One horizontal line and one vertical line but each case
specific analysis needs to perform adequately interpret
the signal obtained, compensating offsets, etc.
Pixel- to- Vector Conversion
A better data retrieving from image vectors and
twice the number of columns pixels are analysed, some
peaks are found successive vertical black lines are low
International Journal of Pure and Applied Mathematics Special Issue
467
resolution images. Each component vector is a
complex number. There are two components are same
column and identical real part and the column index
of pixel matrix. The two consecutive imaginary parts
are quantify the upper and lower limits of vertical
black line. For instance of 2D-vector V=[…; 11+25i;
11+26i; has coordinates meaning that vertical black
line. Fig 4 deals with ECG chart with no axis. A
stretch vector are used to plotting Figure is V =
[10+26i; 10+26i;
Figure 5. ECG signal Data base
4. Disease Detection Using ECG
Pan Tompkins Algorithm
The Pan Tompkins influence to QRS detection as
compared to others. A literature survey signifies the
important algorithm in detecting QRS peak [4]. The
accuracy of Electrocardiogram waveform extraction
plays vital role in better diagnosis of heart related
illnesses. Normal ECG consists of P wave, QRS
complex and T wave. The ECG waves are reflect to
heart’s activity are P wave muscle contraction of atria
and indicates atrial enlargement. Q wave negative
value are supposed to be 45% less than R wave value
[4].
Figure 6. ECG Signal
Amplitude-:
P wave: 0.45mV
R wave: 1.6 mV
Q wave: 45 percent of R wave
T wave: 0.1 to 0.5 mV Duration -:
P-R Interval: 0.14 to 0.40 sec
Q-T Interval:0.45 to 0.44 sec
S-T Interval:0.05 to0.15 sec
P wave Interval:0.11 sec
QRS Interval: 0.09 sec
resolution images. Each component vector is a
complex number. There are two components are same
l part and the column index
of pixel matrix. The two consecutive imaginary parts
are quantify the upper and lower limits of vertical
black line. For instance of 2D-vector V=[…; 11+25i;
11+26i; has coordinates meaning that vertical black
ith ECG chart with no axis. A
stretch vector are used to plotting Figure is V =
ECG signal Data base
Disease Detection Using ECG Data
The Pan Tompkins influence to QRS detection as
literature survey signifies the
important algorithm in detecting QRS peak [4]. The
accuracy of Electrocardiogram waveform extraction
plays vital role in better diagnosis of heart related
illnesses. Normal ECG consists of P wave, QRS
ECG waves are reflect to
heart’s activity are P wave muscle contraction of atria
and indicates atrial enlargement. Q wave negative
value are supposed to be 45% less than R wave value
:
Pan and Tompkins algorithm identifies the QRS
complex based digital analysis slope, amplitude, and
width of ECG data. This algorithm implements to digital
band pass filter. It can reduce false detection and caused
by various types of interference in the ECG signal.
In this algorithm adjusts parameters
adapt the changes in QRS morphology and heart rate.
The following processing steps are given by
• Band-pass filtering.
• Differentiation.
• Squaring.
• Moving window integration.
• Thresholds adjustment.
Figure 7. QRS complex detection.
P wave is rounded peak and occurred QRS
complex. Therefore P wave found based on the location
of QRS complex. The QRS detection algorithms
included all biomedical signal processing textbooks are
introduced Pan and Tompkins. An
algorithm as follows.
Figure 8. Pan-Tompkins beat detection algorithm
Fig 8 shows graphical representation of the basic steps of
algorithm. The ECG signal passes through filtering,
derivation, squaring, and integration phases before
thresholds set and QRS complexes are detected. In ECG
algorithm passes through a low pass and high pass filter
in order to reduce influence of the muscle noise, power
line interference, baseline wander and the T
interference.
Implementation of SVM for QRS
ECG signal is done by LIBSVM software .LIBSVM is
an integrated software package for support vector
Pan and Tompkins algorithm identifies the QRS
mplex based digital analysis slope, amplitude, and
width of ECG data. This algorithm implements to digital
band pass filter. It can reduce false detection and caused
by various types of interference in the ECG signal.
In this algorithm adjusts parameters periodically
adapt the changes in QRS morphology and heart rate.
The following processing steps are given by
QRS complex detection.
P wave is rounded peak and occurred QRS
complex. Therefore P wave found based on the location
of QRS complex. The QRS detection algorithms
included all biomedical signal processing textbooks are
introduced Pan and Tompkins. An overview of the
Tompkins beat detection algorithm
Fig 8 shows graphical representation of the basic steps of
algorithm. The ECG signal passes through filtering,
derivation, squaring, and integration phases before
holds set and QRS complexes are detected. In ECG
algorithm passes through a low pass and high pass filter
in order to reduce influence of the muscle noise, power
line interference, baseline wander and the T-wave
Implementation of SVM for QRS detection in
ECG signal is done by LIBSVM software .LIBSVM is
an integrated software package for support vector
International Journal of Pure and Applied Mathematics Special Issue
468
classification and distribution estimation. A modified
sequential minimal optimization algorithm to perform
training of SVMs. SMO algorithm breaks into the
large quadratic programming (QP) problem in to a
series of smallest possible QP problems. The small
QP problems are solved analytically, which avoids
using a time-consuming numerical QP optimization
problem as an inner loop [8].
Performance: We have used three parameters for
evaluating performance of our algorithm. The
parameters are accuracy, sensitivity, positive
predictive. The parameters are defined using
measures True Positive (TP), True Negative (TN),
False Positive (FP), and False Negative (FN)
True Positive: arrhythmia detection coincides with
decision of physician.
True Negative: both classifier and physician
suggested absence of arrhythmia
False Positive: system labels a healthy case as an
arrhythmia one
False Negative: system labels an arrhythmia as
healthy
Accuracy: Accuracy is the number of correctly
classified cases, and is given by
Accuracy= (TP+TN) / N----------------------- (1)
Total number of cases are N
Sensitivity: Sensitivity refers to the rate of correctly
classified positive. Sensitivity may be referred True
Positive Rate. Sensitivity should be high for a
classifier
Sensitivity = TP / (TP+FN) -------------------- (2)
Positive predictive: Positive predictive is probability
that disease is present when test is positive, which is
by how much amount disease is correctly predicted.
Positive predictive = TP/ (TP+FP) ---- (3)
5. Results
Figure 9.QRS Complex Detection
Figure 10. SVM Classifier
Figure 11.Type of Disease
6. Conclusion
The proposed method extraction and digitization of ECG
signal from various sources. They are thermal ECG
printouts, scanned ECG and captured ECG images are
from devices is proposed. A reasonably accurate
waveform are free from printed character and as tested
through heart rate, QRS width and stability calculations.
We used in this paper MATLAB software for the Pan
and Tompkins QRS detection algorithm for detection of
diseases and related to heart. Notwithstanding fairly
amount of scientific results and proposed work describes
the foundation of efficient tool to generate digital data
signals legated paper charts. In this paper proposed to
low-cost software tool that can be particularly helpful to
scientists and engineers. It can save data as MATLAB
file or as ASCII files and edit or complement the data.
Further work progress are enhance to overall
efficiency of the method and developing a fuzzy based
expert system that assist a doctor in diagnosis and
generating automatic diagnosis reports as well.
Acknowledgement
I would like to express my special thanks of gratitude to
my Supervisor as well as our principal who gave me the
golden opportunity to do this wonderful project on the
topic, which also helped me in doing a lot of Research
and I came to know about so many new things I am
really thankful to them.
References
[1] Vijayaragavan S.P., Karthik B., Kiran T.V.U.,
Sundar Raj M., Robotic surveillance for patient care in
International Journal of Pure and Applied Mathematics Special Issue
469
hospitals, Middle - East Journal of Scientific
Research, v-16, i-12, pp-1820-1824, 2013.
[2] Vijayaragavan S.P., Karthik B., Kiran Kumar
T.V.U., Sundar Raj M., Analysis of chaotic DC-DC
converter using wavelet transform, Middle - East
Journal of Scientific Research, v-16, i-12, pp-1813-
1819, 2013.
[3] Sundararajan M., Optical instrument for
correlative analysis of human ECG and breathing
signal, International Journal of Biomedical
Engineering and Technology, v-6, i-4, pp-350-362,
2011.
[4] Kiran Kumar T.V.U., Karthik B., Improving
network life time using static cluster routing for
wireless sensor networks, Indian Journal of Science
and Technology, v-6, i-SUPPL5, pp-4642-4647,
2013.
[5] Karthik B., Kumar T.K., Dorairangaswamy
M.A., Logashanmugam E., Removal of high density
salt and pepper noise through modified cascaded
filter, Middle - East Journal of Scientific Research, v-
20, i-10, pp-1222-1228, 2014.
[6] Karthik B., Kiran Kumar T.V.U., EMI
developed test methodologies for short duration
noises, Indian Journal of Science and Technology, v-
6, i-SUPPL5, pp-4615-4619, 2013.
[7] Vijayaragavan S.P., Karthik B., Kiran Kumar
T.V.U., Privacy conscious screening framework for
frequently moving objects, Middle - East Journal of
Scientific Research, v-20, i-8, pp-1000-1005, 2014.
[8] Vijayaragavan S.P., Karthik B., Kiran Kumar
T.V.U., A DFIG based wind generation system with
unbalanced stator and grid condition, Middle - East
Journal of Scientific Research, v-20, i-8, pp-913-917,
2014.
[9] Arul Selvi S., Sundararajan M., A combined
framework for routing and channel allocation for
dynamic spectrum sharing using cognitive radio,
International Journal of Applied Engineering
Research, v-11, i-7, pp-4951-4953, 2016.
[10] Arul Selvi S., Sundararajan M., SVM based two
level authentication for primary user emulation attack
detection, Indian Journal of Science and Technology,
v-9, i-29, pp--, 2016.
[11] Kanniga E., Sundararajan M., Kanembedded
control of sub cyclic Ac chopperwith high speed and
low switching losses, Advanced Materials Research,
v-717, i-, pp-579-584, 2013.
[12] Kanniga E., Sundararajan M., Modelling and
characterization of DCO using pass transistors,
Lecture Notes in Electrical Engineering, v-86 LNEE,
i-VOL. 1, pp-451-457, 2011.
[13] Kanniga E., Selvaramarathnam K.,
Sundararajan M., Embedded control using mems
sensor with voice command and CCTV camera,
Indian Journal of Science and Technology, v-6, i-
SUPPL.6, pp-4794-4796, 2013.
[14] Lakshmi C., Ponnavaikko M., Sundararajan M.,
Improved kernel common vector method for face
recognition, 2009 2nd International Conference on
Machine Vision, ICMV 2009, pp-13-17, 2009.
[15] Lakshmi C., Sundararajan M., Manikandan P.,
Hierarchical approach of discriminative common vectors
for bio metric security, 2010 The 2nd International
Conference on Computer and Automation Engineering,
ICCAE 2010, v-2, i-, pp-784-790, 2010.
[16] Venkataganesan K.A., Mohan Kumar R., Brinda
G., The impact of the determinant factors in the career
satisfaction of banking professionals, International
Journal of Pharmacy and Technology, v-8, i-3, pp-
17431-17436, 2016.
[17] Sambantham M.C., Venkatramaraju D., Human
resources management (HRM) practices in multinational
companies with reference to knowledge transfer,
International Journal of Pharmacy and Technology, v-8,
i-3, pp-18565-18571, 2016.
[18] Suganthi S., Senthilkumar C.B., A study on stress
management of the staff of the co operative banks of
Tamil Nadu, India, International Journal of Pharmacy
and Technology, v-6, i-4, pp-7529-7533, 2015.
[19] Thooyamani K.P., Udayakumar R., Khanaa V.,
Cooperative trust management scheme for wireless
sensor networks, World Applied Sciences Journal, v-29,
i-14, pp-253-258, 2014.
[20] Nivethitha J., Brindha G., Management of Non-
Performing Assets in Virudhunagar District Central Co-
Operative Bank-An Overview, Middle - East Journal of
Scientific Research, v-20, i-7, pp-851-855, 2014.
[21] Ramachandran S., Venkatesh S., Performance
measurement and management system-inter company
case Study approach-Tamilnadu, India, Middle - East
Journal of Scientific Research, v-20, i-9, pp-1162-1174,
2014.
[22] Mathew S., Brindha G., Medical tourism – An
avenue to attract foreign patientsin Indian hospital
industry, a study conducted in Chennai city, International
Journal of Applied Engineering Research, v-9, i-22, pp-
7508-7513, 2014.
[23] Mathew S., Brindha G., An empirical study on
competency mapping – A tool for talent management,
International Journal of Applied Engineering Research,
v-9, i-22, pp-7348-7354, 2014.
[24] Mathew S., Brindha G., Quality of work life
among the women it professionals in Chennai city,
International Journal of Applied Engineering Research,
v-9, i-22, pp-7434-7442, 2014.
[25] Mathew S., Brindha G., Level of stress and its
impact on job satisfaction among the employees of Air
India limited, Chennai, International Journal of Applied
Engineering Research, v-9, i-22, pp-7549-7559, 2014.
[26] Brindha G., Emerging trends and issues in
human resource management, Middle - East Journal of
Scientific Research, v-14, i-12, pp-1727-1730, 2013.
International Journal of Pure and Applied Mathematics Special Issue
470