An Approach to Detect CAD with Full Confidence Using Machine Learning Techniques

The main aim of this paper is to detect the Coronary Artery Disease (CAD) from the phonocardiogram (PCG) signals of human heart sounds using different machine learning (ML) techniques. In this proposed method, we separate CAD and non-CAD groups from PCG signals using different ML techniques with proper denoising. As de-noising is very important for PCG signals and ultimately affects the detection result, we first applied various wavelet-based de-noising techniques for non-CAD and CAD type heart sound signals (HSSs) to filter out the high-frequency components present. The Wavelet Scattering Transform (WST) has been used to extract many relevant features from de-noised signals in a single run, strengthening classification. Nowadays, ML plays a vital role in the health sector, and it has also influenced human life in this modern era. Various ML techniques have been investigated in this paper, and a system has been designed for the diagnosis of heart disease, with an emphasis on CAD for alarm generation. Classification of CAD was done by Naive Bayes(NB), K-Nearest Neighbour(KNN), Random Forest (RF), and Support Vector Machine (SVM) classifiers. Recall, precision, F1-Score, and accuracy were all used to evaluate the proposed models. In comparison to other models, the results based on the aforementioned measures show that SVM performed the best. SVM achieved 100% accuracy, 100% precision, 100% recall, and 100% F1-score. The results show that ML models have the ability to correctly identify CAD patients with appropriate denoising.