Deep Learning for Cardiovascular Health: CNN-Based Automated Analysis of PCG Signals

Since cardiovascular diseases (CVDs) continue to be the largest cause of death worldwide, precise and effective diagnostic techniques are essential. Phonocardiography (PCG), which records minute heart sounds, has the potential to identify cardiac defects early. In order to automate the detection of heart abnormalities, this study explores the use of convolutional neural networks (CNNs) in PCG signal analysis. Five different kinds of heart sounds were pre-processed from a large dataset. To extract complex characteristics from PCG signals, a sophisticated CNN architecture was developed that consists of numerous convolutional and pooling layers. The model was able to identify between normal and pathological cardiac sounds with an accuracy of more than 90 %. Techniques for data normalisation, augmentation, and visualisation were used to improve training effectiveness and validate the model. This study provides an ethical and effective method of identifying circulatory irregularities at an early stage, demonstrating the promise of AI in transforming cardiac diagnostics.