Advancing Cardiac Diagnosis via Machine Learning Techniques to Identify Anomalous ECG Patterns

Early detection of cardiac abnormalities through ECG signal analysis is crucial for improving patient outcomes. This study evaluates machine learning techniques—K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Decision Tree (DT), and Random Forest (RF)—to classify ECG patterns into arrhythmia (ARR), congestive heart failure (CHF), and normal sinus rhythm (NSR) efficiently. Key ECG features include the Stress Index (SI), Vagal Sympathetic Effect (VSE), and Respiratory Sinus Arrhythmia (RSA). SI shows high mean values for NSR (3.86), intermediate for ARR (2.68), and low for CHF (1.32), indicating stress variations among these conditions. VSE values are lowest in CHF (0.52), highest in ARR (0.78), and intermediate in NSR (0.68), reflecting differences in autonomic regulation. RSA values demonstrate a similar pattern: NSR at 3.91 ms, CHF at 2.22 ms, and ARR at 5.21 ms, showing varying levels of respiratory influence on heart rate. These numerical distinctions underscore the different physiological impacts of each cardiac condition. Among the classification techniques, Random Forest consistently outperforms others, making it particularly effective for early diagnosis. This comparative analysis highlights the importance of specific ECG features and the performance of RF in accurately distinguishing between ARR, CHF, and NSR, enhancing timely and precise cardiac condition diagnosis.