Heart Murmur Classification in Phonocardiogram Representations Using Convolutional Neural Networks

Heart murmurs are sounds made by rapid blood flow in the heart. Abnormal heart can be a sign of serious conditions such as arrhythmia and cardiovascular diseases. Therefore, murmur classification is crucial for early detection conditions. To this end, we study problem training selected convolutional neural network (CNN) models (such VGGNet ResNet) using various signal representations spectrogram, mel-frequency cepstral coefficient (MFCC), shorttime Fourier transform (STFT)) phonocardiograms public PASCAL CHSC dataset. Our preliminary results show that ResNet outperforms across all metrics representations, consistent with recent published works find literature. Unlike some these works, however, see MFCC STFT general more effective higher test accuracies than spectrogram CNN models. Looking forward, propose to other InceptionV3 Vision Transformer) predict phonocardiogram including STFT, well others like Wigner Ville distribution.