HEART MURMUR CLASSIFICATION OF CARDIAC VALVE DISORDER USING PCG SIGNALS

Project Title

HEART MURMUR CLASSIFICATION OF CARDIAC VALVE DISORDER USING PCG SIGNALS

Project Area of Specialization Artificial IntelligenceProject Summary

Cardiovascular disease is a silent killer and one of the major causes of 17.9 million deaths worldwide each year. The conventional techniques like Auscultation & ECG make it impossible to capture low amplitude heart sounds also known as Murmurs. In addition, there is a serious lack of Cardiac facilities, consultants and departments. This necessitates the development of a methodology that can accurately and timely detect the presence of Murmurs. We develop an algorithm which will help in early detection of heart murmurs. This project covers the application of signal processing techniques on 175 Phonocardiogram (PCG) signals. It consists of pipeline of steps including Pre-processing stage (Wavelet Decomposition & Reconstruction), Segmentation stage (Average Shannon Energy Envelope & Thresholding) for detection of peaks, (MFCCs) is applied for Feature Extraction stage and Classification stage is done by using (NN tool) as a classifier which gives best result reaching to 90% accuracy level.

Project Objectives

The objective of this project is to detect heart murmurs timely and more accurately, facilitate cardiac consultants and departments with a fast and accurate method for detecting murmurs. Diagnosis of congenital cardiac defects is challenging, with some being diagnosed during pregnancy while others are diagnosed after birth during childhood. Prompt diagnosis allows early intervention and best prognosis. It contributes to about 31% of the total deaths happening around the world each year. 75% of these deaths are happening in middle- & low-income countries. Lack of cardiac consultants and departments leads to an alarming situation throughout the world. This necessitates the development of a methodology that can accurately and timely detect the presence of Murmurs. We aim to develop a technique which will help in early detection of heart murmurs. The current goal of this project is to make a system that can detect heart murmurs based on the defined algorithm as we have come to know that it is impossible to catch heart murmurs using techniques like ECG & Auscultation.

Project Implementation Method

The whole system is based on an algorithm for detecting the heart murmurs by applying four different stages as Pre-processing, Segmentation, Feature Extraction and Classification.

• Pre Processing Stage: - It covers the implication of signal, down sampling and normalization is applied on that signal, then the signal is passed from wavelet decomposition and reconstruction stages to Denoise the PCG signal.
• Segmentation Stage: - The pre-processed signal is then bought to the segmentation stage where its Shannon Energy Envelope is obtained and then thresholding is applied for detection of peaks and diastole period of signal.
• Feature Extraction Stage: - Feature extraction is a special form of dimension reduction, which transforms the input data into the set features. Heart sound is an acoustic signal and many techniques used nowadays for human recognition tasks borrow speech recognition techniques. The best and popular choice for feature extraction of acoustic signals is the Mel Frequency Cepstral Coefficients (MFCC) which maps the signal onto a Mel-Scale which is non-linear and mimics the human hearing.
• Classification Stage: - The proposed classification algorithm has two main steps. In the first step, the bad quality recordings (class 0) are detected. In the second step, among the good quality signals, the normal (class -1) and abnormal (class 1) PCGs are classified.

Benefits of the Project

A machine learning model based on NN tool and MFCC, covered normal heart sounds and abnormal heart sounds including AR, MR, MS, and AS. The machine learning model based on NN tool and MFCC achieved 90% sensitivity. This work supports that NN tool as a classifier and MFCC as a feature matrix are widely used for heart sounds classifications, as they have demonstrated their effectiveness, especially if mixtures of features from different domains were employed. The MFCC 13 features coefficients allowed reduction of calculation time and memory that will impact cost of recognition model. This study comprised the size of real heart sounds (175 heart sounds), of them 40 heart sounds were for testing and135 heart sounds were for training. This research did not study simulated heart sounds, while all previous reports used simulated sounds.

1. Service cost per appointment with doctor 
2. Service hours.
3. Response time
4. No need of ECG.
5. An opportunity for the users to use by its own after some guidence from doctor.

Technical Details of Final Deliverable

This paper has  algorithm that implemented in this work allows segmenting of the PCG signal and calculation of different temporal parameters: the durations of heart sounds S1 and S2, duration of heart murmurs, and the duration of cardiac cycle. Heart segmentation should be done, as it is essential for the diagnosis of heart sounds and heart murmur. With this segmentation, we can easily extract the features of each component of the PCG signal. Indeed, the extracted murmur part of the PCG is then deeply analyzed using Mel Frequency Cepstral Coefficient. By this approach the rich information (the high frequencies) contained in the murmur part of the PCG are seen clearly in a way that one can easily distinguish between the most confusing and abnormal PCG signal records.

Final Deliverable of the Project Software SystemType of Industry Medical Technologies Artificial Intelligence(AI)Sustainable Development Goals Good Health and Well-Being for PeopleRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	70000
Electronics sethoscope	Equipment	1	60000	60000
Appointment with doctors	Miscellaneous	10	1000	10000