Comparative Analysis of Machine Learning Techniques for Phonocardiogram-Based Classification of Cardiac Abnormalities

Abstract

Cardiovascular Diseases (CVDs) remain a leading cause of morbidity and mortality worldwide, necessitating early and accurate detection for effective disease management. This work employs advanced signal processing techniques in conjunction with machine learning methodologies to classify normal and particular cardiac conditions—Aortic Stenosis (AS), Mitral Regurgitation (MR), Mitral Stenosis (MS), and Mitral Valve Prolapse (MVP)—using phonocardiogram (PCG) signals. Preprocessing involved denoising using the Discrete Wavelet Transform (DWT) technique with the db8 wavelet and cA2 component, optimizing noise reduction while retaining valuable features for further analysis. Feature extraction was performed using Mel-Frequency Cepstral Coefficients (MFCC) and Mel Power Spectrogram, providing a robust and efficient representation of heart sounds. Two machine learning models—Deep Neural Network (DNN) and Convolutional Neural Network (CNN)—were used to assess the extracted features. With three hidden layers and 80% of the dataset used for training, the DNN model produced 90%±0.37 accuracy, 89% sensitivity, and 91% specificity. On the other hand, the CNN model, which consists of two fully connected layers and two convolutional layers with max pooling, performed by achieving 96%±0.38 accuracy, 95% sensitivity, and 95% specificity. These results underscore DNN’s enhanced capability in handling complex PCG data and reducing false negatives. This comprehensive study addresses multiple cardiac abnormalities, surpassing previous research that often focuses on a single condition or model. The findings highlight the potential of combining advanced signal processing with deep learning techniques to improve the timely and accurate identification of cardiac abnormalities. Future research will explore additional feature extraction methods and larger datasets to further enhance classification performance. This work significantly contributes to the field of biomedical engineering, offering a framework to improve patient outcomes through advanced diagnostic techniques.