Abstract

A Phonocardiogram (PCG) signal is the graphical representation of the heart sound recording that is obtained by placing a stethoscope on the patient’s chest. Since the PCG signal is highly correlated with the Electrocardiogram (ECG), it is imperative to analyze the relationship between the two waveforms.

This work aims to translate the PCG waveform into an equivalent (and meaningful) ECG waveform and vice versa. Two methods will be studied. The first method involves the use of so-called generative adversarial networks (GAN), while the second method involves the computation of the discrete cosine transform (DCT) of both PCG and ECG signals, followed by learning the regression between the two DCT sequences. A natural benefit of studying this relationship is to get insights into cardio disease from PCG data and to get insights into heart valve diseases using the ECG data.  

Deliverables

A research paper (conference/journal).