Multi-Class Cardiac Diagnostic Decision Support System Based On Phonocardiogram Signal

Abstract

A Phonocardiogram (PCG) signal is the representation of signal like murmurs and sounds which made by the vibrations caused for the period of cardiac cycle . Where the heart beat is recorded using a lowcost small handled digital device called as stethoscope . By using this device it provide information the heart rate , intensity , tone , frequency , quality and the location of the various components of the cardiac sound . Due to these characters , phonocardiogram signals ca detect the heart status at the early state and go for the treatment . Diagnoses at the early stage is the only way to decrease the death rate due to cardiac vascular diseases (CVD) . There are many invasive and non invasive method to diagnoses the cardiac vascular diseases . The facilities are no available in the low and middle income areas where the lack of availability of facilities or lack of money the facilities are no easy available and it case death . In previous studies , it uses convolutional neural network (convnet) , which is trained by hybrid constant – Q transform (HCQT) for heart beat sound classification and most studied architecture . Constant Q – transform (CQT) , variable – Q transform (VQT) and hybrid constant Q-transform which is extracted from phonocardiogram signals as the acoustic features , which includes the domains of Mel Frequency Cepstral Coefficients ( MFCCs ) where audio or speech signal processing . In the proposed system convolutional neural network & CQT, Variable-Q Transform (VQT), and HCQT are extracted from each phonocardiogram signal as the acoustic features, including the dominant MFCC features, feed into five-layer regularized convnet like convolution layer , pooling layer and dense layer. After analyzing the literature in the same domain, it can be stated that this is the first time HCQT is being utilized for PCG signals. HCQT is more effective than standard CQT and other variants. Also, the accuracies of the system proposed in this work on the validation datasets are 94% in multi-class classification, which outperforms the proposed work relative to other models significantly