2
FIELD OF THE INVENTION
The present invention relates to an integrated system for diagnosis of valvular
heart diseases. The present invention further relates to a method for diagnosis of
valvular heart diseases.
BACKGROUND OF INVENTION
Auscultation, the technique of listening to heart sounds, remains a primary
detection tool for diagnosing heart valve disorder. The poor sensitivity of human
ears in the low frequency range of the heart sounds makes this task more
difficult. The prevailing work mostly involves data acquisition and segmentation
of phonocardiogram (PCG) signal to produce one cardiac cycle. It usually uses
auxiliary input like electrocardiogram signal for this purpose. Work has been
reported in journals that extracts different kind of features from PCG signal and
identification of up to six pathological cases with 97 percent of accuracy on a
sample size of 28 persons. Apparatus are available from Littman, Meditron etc.
only for acquisition of data and its display. Biosignetics Corporation, USA has
applied for US patent for their 'phonocardiograph monitor' (www.bsignetics.com)
which grades a murmur for further diagnosis of heart disorder.
US patent no. US 5687738 discloses an apparatus for assisting in the analysis of
heart sounds, ECG, and respiratory data. Once digitized, this data is processed
and analyzed to determine timing relationships between the three signals,
frequency (or pitch) of sounds, and dependence or non-dependence of sounds
on ECG and respiratory phase. The user of the apparatus inputs the place of

3
detection and the maneuver being performed. Raw phonocardiogram data is
displayed. The user inputs the number of beats and the frames per second to
display. Fast Fourier transformed and signals averaged data are displayed, anj
phase sensitive and non-phase sensitive sounds are extracted. A lesion fitting
algorithm suggests diagnoses and possible further maneuvers to perform. The
data obtained is compared to a historical patient data.
Auscultation, the technique of listening to heart sound, is used as a primary
detection tool for diagnosis of heart valve disorders since invention of
stethoscope in 1816 by Lannec [1]. Certain heart diseases are best detected onlv
by auscultation [2]. The collection of sonic waves from the surface of the body,
as performed with the stethoscope, continues to provide an important source of
clinical information. Together with the overall bedside examination, this is not
only cost-effective but also non-replaceable by alternative technological methods
[3]. Moreover, echocardiography is not required for all patients with systolic
murmurs [3], [4]. The poor sensitivity of human ears in the low frequency range
of the heart sounds have hampered the physician's ability to accurately collect,
reproduce, and document acoustic data obtained in this fashion. It has been
reported that a disturbing percentage of medical graduates cannot properly
diagnose heart conditions using stethoscope [5]. Paradoxically, phonocardiogram
(PCG), the digitized recording of heart sound is proven to be very useful in the
description and understanding of heart sounds as it provides a visual display of
the recorded waveform [6], [7] and allows computer aided signal processing and
pattern recognition techniques to characterize them. PCG has a high potential to
detect cardiac abnormalities because it provides the valuable information of
functioning of heart valves and the hemodynamics of the heart [6],[8],[9]. Unlike

4
echocardiography, the non-invasive cardiac auscultation is simple, cost-effective
and with proper signal processing and pattern recognition tools can emerge as
an important device for primary detection of heart valve disorders.
In gist, in the prior art apparatus, the three signals (heart sounds, ECG, and
respiratory data) are digitally processed and displayed visually to allow the user
to visually interpret heart sounds based on their timing compared to other heart
sounds, to respiration and to ECG signals.
US patent no. US 4905706 describes a method and apparatus for detection of
coronary artery disease. The apparatus records and analyzes that portion of the
phonocardiogram lying between about 100 to 600 Hz. An electrocardiogram is
also recorded and examined in order to pinpoint the diastolic window of PCG
data. This window of data is subject to autocorrelation analysis and spectral
analysis, resulting in a partial correlation coefficient index and a power density
index. A linear combination of these two indices results in a Cardiac Screening
Index which is indicative of the presence or absence of coronary artery disease.
In gist, this prior art invention claims a method for detection of coronary artery
disease from phonocardiogram and electrocardiogram signal coupled to a
digitizer and processor, which analyzes the acoustic energy of the heart in the
frequency range of 100 to 600 Hertz.

5
OBJECTS OF THE INVENTION
It is therefore an object of the invention to propose an integrated system to
acquire and process digitized heart sound signals for diagnosis of valvular heart
diseases, which eliminates the disadvantages of prior art.
Another object of the invention is to propose an integrated system to acquire
and process digitized heart sound signals for diagnosis of valvular heart diseases,
which is a non-invasive phonocardiogram based, easy to implement and cost -
effective system.
A further object of the invention is to propose an integrated system to acquire
and process digitized heart sound signals for diagnosis of valvular heart diseases,
which incorporates an automatic segmentation unit which is more accurate.
A still another object of the invention is to propose an integrated system to
acquire and process digitized heart sound signals for diagnosis of valvular heai t
diseases, which incorporates an artificial neural network which acts as a classifier
to discriminate the heart sound signal.
Yet another object of the invention is to propose an integrated system to acquire
and process digitized heart sound signals for diagnosis of valvular heart diseases,
which allows capturing patient's heart sound for unlimited time for accurate
diagnosis.

6
A still further object of the invention is to propose an integrated system to
acquire and process digitized heart sound signals for diagnosis of valvular heart
diseases, which allows repetitive play back mode and provides zooming facilities
of the recorded sound for better diagnosis.
Yet a further object of the invention is to propose an integrated system to
acquire and process digitized heart sound signals for diagnosis of valvular heart
diseases, which provides both automatic and manual mode disease detection
facility.
SUMMARY OF INVENTION
Accordingly there is provided an integrated system adaptable for diagnosis of
valvular heart diseases comprising a data acquisition subsystem to non-invasively
capture heart sound from a subject's body, record the sound signals, and
generate digitally processed phonocardiogram signal; a data compression and
decompression subsystem to receive the recorded PCG signal from the data
acquisition subsystem for compression of the recorded PCG signals in a wavelet
based decomposition means and further compression in an Adaptive Differential
Pulse Code Modulation encoder to quantize a difference signal, the difference
signal being the difference between a predicted signal value and the actual signal
value, of the signal; and a data decompression unit for receiving the quantized
value from the ADPCM to perform an inverse quantization and subtract the value
from the predicted signal value, and produce a decoded signal; a recording and
display subsystem for storing the compressed data from the ADPCM in a secured
format (.hsa) for disallowing unauthorized access, and

7
displaying the PCG data in online/offline mode including diagnosing and report
generation; and a decision making subsystem having means for segmentation,
feature extraction and classification of data received from the data compression
and decompression subsystem and allowing display of the identification results
on heart valve disease in the recording and display subsystem.
BRIEF DESCRIPTION OF THE ACOMPANYING DRAWINGS
Fig. 1 - a schematic representation of the data acquisition subsystem of th?
integrated system of the invention.
Fig. 2 - a block diagram of a compression and decompression subsystem of the
integrated system of the invention.
Fig. 3 - a flow diagram showing the operation of an ADPCM encoder unit of the
data compression and decompression subsystem of the integrated system of the
invention.
Fig. 4 - a flow diagram showing the operation of an ADPCM decoder unit of the
data compression and decompression subsystem of the integrated system of the
invention.
Fig. 5 - a flow diagram of the recording and display subsystem of the integrated
system.

8
Fig. 6 - shows a flow-diagram of the operation of the Decision making
subsystem of the integrated system of the invention.
Fig. 7 - a pictorial view of the front end of the integrated system including an
input dialogue box according to the invention.
Fig. 8 - shows the integrated system of the invention in a data recording mode
which represents normal heart sound detection in relation to a first-test case.
Fig. 9 - shows selection of previously recorded data and self test for Aortic
Insufficiency in a process of testing of the integrated system of the invention.
Fig. 10 - shows report generation and training of a physician on the integrated
system of the invention.
Fig. 11 - shows recorded data-representing normal heart sound in relation to a
second test case in an integrated system of the invention.
DETAILED DESCRIPTION OF THE INVENTION
The present invention is directed to a system for processing digitized heart sound
signals for monitoring heart valve disorder, comprising of four subsystems; a
data acquisition subsystem, a data compression and decompression subsystem;
a recording and display subsystem; and a decision making subsystem. The
description of each subsystem is given below. The invention is to serve as an
assisting tool for diagnosis of heart valve related diseases.

9
DATA ACQUISITION SUBSYSTEM (PASS)
The data acquisition subsystem (DASS) is used to acquire the heart sound signal
from a subject. The input of this subsystem (DASS) is the subject's heart sound
signal and output is 16 bit PCM coded digitized heart sound. Fig. 1 is the
functional block diagram showing the main features of the data acquisition
subsystem (DASS) of the inventive system. Heart sounds are captured by the
DASS for analysis of the sound signals and analyzed in next part of the invention.
From the DASS system, the heart sound signal of the subject is recorded for
analysis and detection of possible valvular heart diseases. The recorded sound is
also displayed for visualization and can be heard from speaker output. The
followings are the detailed descriptions of different blocks.
A stethoscope (BO) is placed on the chest of the subject. The stethoscope (BO) is
used as a sound detector, which is a conventional microphone having sound
shielding properties. The stethoscope (BO) which is used in diaphragm mode
picks up the internal acoustical activity of the heart. The MDF dual head,
Mukherjee or KRL stethoscope can be used for capturing the phonocardiogram
signal. A microphone (Bl) with 25 KHz audio band can be used for collecting the
heart sounds of the subject. A pre-amplifier (B2) is used for the amplification of
the collected heart sound signal by the microphones (Bl). The known BC 14813,
NPN transistors are used for developing the pre-amplifier (B2). The output from
the pre-amplifier (B2) is passed through two consecutive low pass filters (B3,
B4). The important information utilized by this device is contained in heart
sounds whose frequency is in lower range. Therefore, a 2 KHz PCG signal is
collected after using two low pass filters. The first low pass filter (B3) with 10

10
KHz cutoff frequency is used after the Pre-amplifier (B2) and followed by the
second filter (B4) whose cutoff frequency is 2 KHz. Filters output is connected to
a the frequency gain amplifier (B5) which amplifies the 0-2 KHz range heart
sound signal. A buffer (B6) is used to record the heart sound signal for further
analysis. The captured heart sound signal can be heard by using a speaker
through an audio amplifier (B7).
DATA COMPRESSION AND DECOMPRESSION SUBSYSTEM (DCSS)
The data compression and decompression subsystem (DCSS) consists of a Data
compression unit (DCU) based on wavelet decomposition and reconstruction
(100) followed by a ADPCM encoder (ECR) and a decompression unit (DDCU),
based on a ADPCM decoder (DCR). The input of the compression unit is raw
digitized PCG signal and output is compressed form of the 16 bit PCG signal.
After compression, the PCG signal is encrypted (101) and stored (102) in private
.hsa format. For further analysis, the stored encrypted signal is first decrypted
(103) and then decompressed (104) by ADPCM decoder (DCR) unit and the
output digitized 16 bit signal is fed to the subsequent subsystems.
DATA COMPRESSION UNIT f DCU)
A data compression subsystem (DCU) seeks to minimize the number of bits
stored by reducing the redundancy present in the PCG signal coming from data
acquisition subsystem (Fig. 1). A novel wavelet based Adaptive Differential Puls^
Code Modulation (ADPCM) [10] has been used in the invented system for storing
the heart sound data. Fig. 2 represents the block diagram of the adapted

11
compression subsystem where the PCG data from Fig. 1 is, first compressed
(100) by the wavelet based decomposition unit and then it is further compressed
by the ADPCM encoder (ECR) (Fig. 3).
The recorded PCG signal (from Fig. 1) is first decomposed by Daubechies-2
wavelet in six levels. It is seen that 1st level detail coefficient contains mostly
high frequency noise, the 2nd, 3rd, 4th, 5th and 6th level detail coefficients and 1st
level approximation coefficients are used to reconstruct (100) the original signal.
This reconstructed signal is further input to the ADPCM encoder (ECR) (Fig. 3).
ADPCM which is a digital compression unit (DCU) is a waveform codec that can
be used to code sound signals. ADPCM is simpler than advanced low bit-rate
voice coding units and doesn't require as heavy calculations, which means
encoding and decoding can be done in a relatively short time.
ADPCM is used to compress with an inherent flow rate of 64 Kbit/s. When
encoded at the highest compression ratio, using only 2 bits to code the ADPCM
signal, the flow rate is reduced to 16 Kbit/s, i.e. 25% of the original. Using 4-bit
coding, the flow rate is 32 Kbit/s, i.e. 50% of the original, and the quality of the
signal is acceptable for this applications.
The principle of ADPCM is to predict the current signal value from previous
values and to transmit only the difference between the real and the predicted
value. In plain Pulse-Code Modulation (PCM) the real or actual signal value would
be transmitted. In ADPCM the difference between the predicted signal value and

12
the actual signal value is usually quite small, which means it can be represented
using fewer bits than the corresponding PCM value. Depending on the desired
quality and compression ratio, a difference signal is quantized using 4, 8, 16 or
32 levels. A flow-diagram on of an ADPCM encoder unit (ECR) is shown in Fig. 3.
The signal is further encrypted and stored in a private format .hsa which can be
played in a media player but does not allow access and hence manipulation of
raw data is not possible. This ensures privacy of the data as well as reduction of
storage space.
DATA DECOMPRESSION UNIT (DDCU)
The decoder unit shown in Fig. 4 takes the quantized value, performs an inverse
quantization, and subtracts the result from the predicted signal to get the
decoded signal for further analysis in subsequent subsystems.
RECORDING AND DISPLAY SUBSYSTEM (RDSS)
Fig. 5 is a flow diagram of recording and display subsystem (RDSS), for storing
PCG data after capturing, displaying in online/offline mode, diagnosing and can
be used to linked to Hospital Management System and Patient database and
Report generation/archiving/printing facility. This subsystem is used to display
the acquired digitized PCG data after capturing through the data acquisition
subsystem (DASS), and computer sound card. The compressed data is stored by
this subsystem in new private .hsa format after compressing through data
compression unit (DUC). The stored compressed data is first decompressed and
then processed for further analysis through decision making subsystem (DMSS).

13
The different outcomes of the decision making subsystem (DMSS) are fed to this
subsystem (RDSS) for display and report generation in auto/manual mode.
As indicated at block SC-7, the output from sound card is continually displayed
and when a stop recording event occurs, the offline display is enabled to store
the phonocardiogram data into .hsa format as well as playback mode, unlimited
zooming and frequency spectrum display modes are enabled. The
phonocardiogram data can be recorded for an unlimited time interval (limited
only by virtual memory which is sufficiently large for all practical purposes), and
when the recording times out (block SC-7) storage of the data in the hard-disk
terminates (in our novel .hsa format as discussed before) with control returning
to the input of Fig. 1. A physician may select to listen the recorded heart sounds
entirely or selected portion and diagnose by Auto or Manual Mode. For example,
the test at block SC-8 is based upon the Auto Mode diagnosis after decision
processing made by Fig. 6. The block SC-8 display the % of diagnosis decision
accuracy along with one segmented cycle of the recorded sound, its spectrogram
and frequency spectrum. If the physician is not satisfied with the decision he can
consult with Manual Mode (block SC-5) and compare the result with normal and
33 different pathological heart sounds. After proper decision, the testing report
can be generated by block SC-3 and print of that report can be generated by
block SC-4. The patient's details can be stored by block SC-1 and can update the
details of the existing patient information also.
The block SC-2 is used to store the discharge information of the patient. Block
SC-5 is used to train physicians and technicians on different kind of
phonocardiogram sounds normal as well as pathological with time domain

14
display, playback facility, unlimited zooming, frequency spectrum and
spectrogram display. A self-test facility in block SC-6 is used to show the % of
diagnosis decision accuracy of 9 common pathological case as well as normal
sounds for the auto mode detection. While the invention has been illustrated and
described in detail in the drawings and foregoing description, the same is to be
considered as illustrative and not restrictive in character, it being understood th?t
only the preferred embodiment has been shown and described and that all
changes and modifications that come within the spirit of the invention are
desired to be protected.
DECISION MAKING SUBSYSTEM (DMSS)
The decision making subsystem (DMSS) consists of segmentation, feature
extraction and classifier units for identification of heart valve diseases. The
stored compressed signal is first decompressed by decompression unit (DDCU)
and the decompressed signal is fed to this subsystem (DMSS) and processed
outcomes are displayed by recording and display subsystem (RDSS).
The complete cardiac cycle can be extracted from different established
segmentation methods [11, 12]. In this invention, a novel segmentation block
has been proposed. The segmentation of stored PCG in Fig. 5 is primarily based
on the use of frequency content present in the signal, calculation of energy in
time windows and timing relations of signal components. Various medical domain
features e.g. normal split-sound duration, frequency content of S1 and S2 etc.
are exploited in this subsystem.

15
According to the invention, Wavelet based feature extraction is applied as
described in [13]-[17] to obtain the features of the segmented PCG signals. Use
has been made of a wavelet based feature extraction block mentioned in [14],
[16] where Daubechies-2 wavelet is used for determining the wavelet
coefficients. Wavelet detail coefficients and approximation coefficient at the sixth
level for normal and 9 pathological subjects are calculated. The signal formed by
wavelet detail coefficients at the second decomposition is regrouped into 32 sub-
windows. The elements of the feature vectors are formed by the powers of the
signal (32 values) within these sub-windows.
Artificial Neural Network (ANN) unit employed for heart sound modeling uses
back propagation multi-layer perceptron (BP-MLP) mechanism. In this case, the
problem is to classify the feature vectors into several heart sound classes. The
network unit consists of an input layer, one hidden layer and an output layer.
Here, the number of nodes in the input layer equals to 32 feature dimensions
whereas number of nodes in output layer is always the number of heart sounds
to be classified. The number of nodes in the hidden layer is chosen 55 which
show a superior performance for this subsystem.
The present invention thus provides an integrated system to acquire and process
digitized heart sound signal for diagnosis of valvular heart diseases which
comprises
• a data acquisition subsystem to capture and record the heart sound from
subject's body and generates digitally processed PCG signal.

16
• a data compression and decompression subsystem to compress the
recorded heart sounds by wavelet and ADPCM unit followed by encryption
unit to store the data in a newly developed .hsa format which reduce
storage space and ensure privacy of the PCG data, and the stored data
comprising heart sound is decrypted and decompressed for further
analysis,
• a recording and display subsystem to store PCG data after capturing and
compression, displaying the data in online/offline mode, diagnosing and
report generation/archiving/printing facility; and
• a decision making subsystem to identify heart valve diseases with the help
of segmentation unit, feature extraction and modeling unit.
• The present invention further provides a method for detecting valvular
heart disease suitable for primary health diagnosis.
• The invention provides an integrated apparatus which does not require
invasion of the body by instruments or by acoustic waves or
electromagnetic radiation.
• According to the inventive method, PCG data is first compressed by
wavelet unit and then further compressed by ADPCM unit. Therefore,
more compressed data is produced which reduces the storage space. The
recorded PCG signal is first decomposed by Daubechies-2 wavelet in six

17
levels. Since 1st level detail coefficient contains mostly high frequency
noise, the 2nd, 3rd, 4th, 5th and 6th level detail coefficients and 1st level
approximation coefficients are used to reconstruct the original signal.
Therefore, a compressed data is found by rejecting 1st level detail
coefficient of the PCG signal. This reconstructed signal is further input to
the ADPCM encoder unit.
• The compressed data is stored in a newly developed .hsa format to
ensure privacy of the data.
• The decision making subsystem uses a novel segmentation unit. In
contrast to the prior art segmentation methods which use
electrocardiography (ECG) signal as a continuous auxiliary input in a
complex instrumentation setup, the invented method performs an
automatic segmentation that does not require any auxiliary signal. The
method extensively utilizes biomedical domain features for reduction of
time and computational complexities and is more accurate. The
segmentation unit uses auto correlation block to calculate the duration of
recoded cardiac cycles. The PCG signal is wrapped with the length of one
cardiac cycle to get accurate length of cycle. Thereafter, with help of
other downstream units as illustrated discussed in Fig. 6, the S1 and S2
are detected for one cardiac cycle. The exact boundary of the S1 and S2 is
calculated by segmentation unit to get accurate cycle. Segmentation using
end correction unit describes the boundary of the S1 and S2. This
boundary detection unit utilizes the biomedical features and time-domain
relationship of the signal to reduce the complexity of the invented system.

18
• The invention thus provides a simple, cost-effective integrated apparatus
by which non-medical personnel may screen patients without the
supervision of professional physicians. The placement of stethoscope and
other important details are clearly defined using suitable diagram.
• The foregoing objects are achieved in a device for detection of valvular
heart disease comprising a phonocardiogram coupled to a digitizer and PC
which analyzes the acoustic signal energy of the heart. This energy is
compared to certain statistical characteristics of the waveform, yielding a
final screening index indicative of heart valve disease.
ADVANTAGES
1. The system of the invention constitutes a phonocardiogram based heart
valve disorder system which is easy to implement and cost-effective.
2. If patients with functional systolic murmur can be identified and not
routinely referred for echocardiography, substantial cost savings can be
realized [3][4].
3. The data acquisition subsystem by which the heart sound signal of subject
is recorded for further analysis is low cost device, which is comprised only
of microphone, filter and amplifiers.

19
4. The compressed form of patient's recorded data is stored to save disk's
space. The stored data is decompressed to get back its original stage
when it is necessary. The new data format designed ensures privacy of
data.
5. The system is configured by incorporating an automatic segmentation unit
that extract cardiac cycles from continuous phonocardiogram signal and is
more accurate than any other existing segmentation units that does not
use auxiliary input like ECG and avoids complex instrumentation setup.
6. Represent heart sound in an understandable plot format that allows clear
differentiation of S1, S2 and murmurs.
7. The time duration of a cycle is also shown by auto-correlation block of
segmentation unit which in turn shows the heart rate of the subject.
8. The system uses wavelet based feature extraction unit that is popular in
extracting feature vectors from heart sound signal because of its ability to
characterize time-frequency information which is important in this context.
9. Artificial Neural Network (ANN) is used in this system as a classifier to
discriminate the heart sound signal using these features. The ANN can be
used to generate likelihood-like scores that are discriminative in the state
level, can be easily implemented in hardware platform for its simple
structure, has the ability to approximate functions and automatic similarity

20
based generalization property, and complex class distributed features can
be easily mapped by the ANN.
10. Capturing of a patient heart sound can be done for unlimited time.
11. Online display facility with storing, playback, zooming and frequency
analysis.
12. Repetitive playback mode and unlimited zooming facility of the recorded
sound or a part of it.
13. Self Test, Automatic and Manual mode disease detection facility is
available.
14. Heart disease detection facility is incorporated which compares the normal
and 33 different pathological heart sounds.
15.Allows examining precise frequency spectrum, spectrogram and
segmented cycle at any moment in time.
16. It is used to train Physicians/Technicians on different kind of
Phonocardiogram sounds, normal as well as pathological.
17. It can be linked to Hospital Management System and Patient database
and Report generation/archiving/printing facility.

21
18. Enhanced license protection features, cannot be copied to other
computers without a license, registration and license key.
TESTING
Figure 7 shows the front end of the integrated system including an input
dialogue box on patient information. The system can diagnose 9 pathological
cases and normal heart sounds. The database is divided in two sections each of
which consists of 320 cycles from 40 volunteers. One set is used for training the
network and the other set is used for testing. The testing shows the recognition
accuracy above 95 % for normal and 9 commonly occurring pathological heart
sounds for auto-mode diagnosis. The system also shows the visual display of
recorded heart sounds and its frequency spectrum through Fast Fourier
Transform (FFT) of this sound. A segmented cycle of recorded waveform and its
spectrogram (both time and frequency information) and spectrum can be seen
which gives important information to physicians for manual diagnosis. The
system is tested on a PC with following specification:
Pentium III 600 MHz or higher or AMD 600 MHz or higher, 128 Mb of RAM (512
MB preferred), Internal or external sound card (Realtek AC97 Audio driver,
SoundMax Integrated digital audio driver, etc.), Windows XP Home or
Professional or Windows 2000 Professional. Figures 8 to 11 represent the
different visual outputs of the system including identifying pathological cases in
auto mode, manual mode and training mode.

22
REFERENCES
[1] Ibrahim R. Hanna, Mark E. Silverman, "A history of cardiac auscultation and
some of its contributors", The American Journal of Cardiology, vol. 90, pp. 259-
267, Aug. 1, 2002.
[2] J. R. Bender, "Yale University School of Medicine Heart Book", New York:
William Morrow and Company, Inc., 1992, ch. 13, pp. 167-175.
[3] M. E. Tavel, "Cardiac Auscultation: A Glorious Past—But Does It Have a
Future?", circulation, vol. 93, pp. 1250-1253,1996.
[4] C. Shub, "Echocardiography or auscultation? How to evaluate systolic
murmurs?", Can. Fam. Physician, vol. 49, pp. 163-167, 2003.
[5] S. Mangione, L. Nieman, "Cardiac auscultatory skills of internal medicine and
family practice trainees", Journal of the American Medical Association, vol. 278,
pp. 717-722,1997.
[6] L. G. Durand, P. Pibarot, "Digital signal processing of the phonocardiogram:
review of the most recent advancements", Critical Reviews in Biomedical
Engineering, vol. 22, no. (3/4), pp. 163-219,1995.
[7] S. Lukkarinen, A. Nopanen, K. Sikio, A. Angerla, "A new phonocardiographic
recording system", Computers in Cardiology, vol. 27, pp. 117-120,1997.

23
[8] Leslie Cromwell, Fred J. Weibell, Erich A. Pfeiffer, "Biomedical
Instrumentation and Measurements", 2nd ed, June, 2002, PHI Publication, Ch. 6:
169-172.
[9] Rangaraj M. Rangayyan, "Biomedical Signal Analysis", 2002, IEEE Press, John
Willy & Sons Inc., Ch. 1:34-38.
[10] A. M. Kondoz, "Digital speech", (John Wiley & Sons Ltd.), Second ed., 2004.
[11] H. Liang, S. Lukkarinen, I. Hartimo, "Heart Sound Segmentation Algorithm
based on Heart Sound Envelogram", Computers in Cardiology, vol. 24, pp. 105-
108,1997.
[12] H. Liang, S. Lukkarinen, I. Hartimo, "A heart sound segmentation algorithm
using wavelet decomposition and reconstruction", in Proceedings of the 19th
Annual International Conference of the Engineering in Medicine and Biology
society, IEEE, vol. 4, pp. 1630-1633, 30 Oct.-2 Nov., 1997.
[13] Ian Cathers, "Neural network assisted cardiac auscultation", Artificial
Intelligence in Medicine, vol. 7, pp. 53-66,1995.
[14] T. Olmez, Z. Dokur, "Classification of heart sounds using an artificial neural
network", Pattern Recognition Letters, vol. 24, pp. 617-629, Jan., 2003.

24
[15] T. R. Reed, N. E. Reed, P. Fritzson, "Heart sound analysis for symptom
detection and computer-aided diagnosis", Simulation Modelling Practice and
Theory, vol. 12, pp. 129-146, 2004.
[16] C. N. Gupta, R. Palaniappan, S. Swaninathan, S. M. Krishnan, "Neural
Network classification of homo-morphic segmented heart sounds", Applied Soft
Computing, vol. 7, pp. 286-297, 2007.
[17] S. Omran, M. Tayel, "A heart sound segmentation and feature extraction
algorithm using wavelets", in First international symposium on control,
communication and signal processing, pp. 235-238, 2004.

25
WE CLAIM
1. An integrated system adaptable for diagnosis of valvular heart diseases
comprising:
- a data acquisition subsystem (DASS) to non-invasively capture
heart sound from a subject's body, record the sound signals, and
generate digitally processed phonocardiogram (PCG) signal;
- a data compression and decompression subsystem (DCSS) to
receive the recorded PCG signal from the data acquisition
subsystem (DASS) for compression of the recorded PCG signals in a
wavelet based decomposition means and further compression in an
Adaptive Differential Pulse Code Modulation (ADPCM) encoder
(ECR) to quantize a difference signal, the difference signal being
the difference between a predicted signal value and the actual
signal value; and a data decompression unit (DDCU) for receiving
the quantized value from the ADPCM (ECR) to perform an inverse
quantization and subtract the value from the predicted signal value,
and produce a decoded signal;
- a recording and display subsystem (RDSS) for storing the
compressed data from the ADPCM in a secured format (.hsa) for
disallowing unauthorized access, and displaying the PCG data in

26
online/offline mode including diagnosing and report generation;
and
- a decision making subsystem (DMSS) having means for
segmentation, feature extraction and classification of data received
from the data compression and decompression subsystem (DCSS)
and allowing display of the identification results on heart valve
disease in the recording and display subsystem (RDSS).
2. The system as claimed in claim 1, wherein the data acquisition subsystem
(DASS) comprising a sound detector (B1) for non-invasively capturing
heart sound signal of a subject; a pre-amplifier (B2) having two low pass
filters (B3, B4); a frequency gain amplifier (B5) for amplifying the sound
signal at different frequency ranges; and buffer (B6) for recording the
captured heart-sound signal.
3. The system as claimed in claim 1, wherein the data compression and
decompression subsystem (DCSS) comprises a data compression unit
(DCU) receiving a raw digitized PCG-signal from the data acquisition
subsystem (DASS) and outputting a compressed form of 16 bit signal; a
ADPCM encoder (ECR), encryption and storage of data; decryption of the
encrypted data and a data decompression unit (DDCU), for further
decompression.

27
4. The system as claimed in claim 1, wherein the decision making subsystem
(DMSS) comprises a segmentation means; a feature extraction means,
and a classifier means, wherein the segmentation means carrying out the
segmentation of stored PCG-data by calculation of energy in time windows
and timing relations of the signal components, wherein the feature
extraction means determines the wavelet based coefficient, wavelet detail
coefficients and approximation coefficient, and wherein the classifier
means being employed for heart sound modeling for identification of heart
valve diseases.
5. An integrated system adaptable for diagnosis of valvular heart diseases as
herein described and illustrated with reference to the accompanying
drawings.

Accordingly there is provided an integrated system adaptable for diagnosis of
valvular heart diseases comprising a data acquisition subsystem to non-invasively
capture heart sound from a subject's body, record the sound signals, and
generate digitally processed phonocardiogram signal; a data compression and
decompression subsystem to receive the recorded PCG signal from the data
acquisition subsystem for compression of the recorded PCG signals in a wavelet
based decomposition means and further compression in an Adaptive Differential
Pulse Code Modulation encoder to quantize a difference signal, the difference
signal being the difference between a predicted signal value and the actual signal
value, of the signal; and a data decompression unit for receiving the quantized
value from the ADPCM to perform an inverse quantization and subtract the value
from the predicted signal value, and produce a decoded signal; a recording and
display subsystem for storing the compressed data from the ADPCM in a secured
format (.hsa) for disallowing unauthorized access, and
displaying the PCG data in online/offline mode including diagnosing and report
generation; and a decision making subsystem having means for segmentation,
feature extraction and classification of data received from the data compression
and decompression subsystem and allowing display of the identification results
on heart valve disease in the recording and display subsystem.