ECG Multi Class Classification Using Machine Learning Techniques

This paper presents a project focused on utilizing Artificial Intelligence (AI) tools to improve the process of diagnosing heart diseases. The research indicates that 10 to 15 percent of Pacific Islanders are diagnosed with at least one form of heart disease, leading to around 20,000 deaths annually. The proposed project uses the Physio net database and ECG signals of 162 patients to design a multi-class classification method that accurately recognizes different patterns under 3 classes, namely, Arrhythmia (ARR), Congestive Heart Failure (CHF), and Normal Sinus Rhythm (NSR). The study utilizes two feature extraction methods, Continuous Wavelet Transform, and Wavelet Scattering, to extract the principal characteristics from the ECG data. MATLAB Software is used to train three models, an AlexNet Model, an SVM Model, and an LSTM Model, to diagnose cardiovascular diseases and their severity. The results of the different classification methods showed that the SVM Model had the best performance with a classification accuracy of 98%. This project offers a dependable and effective diagnostic tool for the diagnosis of heart diseases with a minimized risk of human error. Additionally, it has the potential to serve as a valuable resource for future studies in the medical field aimed at enhancing cardiovascular disease diagnosis and treatment.