Automated Identification of Cardiac Anomalies from Electrocardiographic Signals

Over recent decades, cardiovascular diseases (CVDs) have become the foremost cause of mortality in both industrialized and emerging countries. Timely diagnosis and consistent medical monitoring are critical in decreasing death rates linked to these illnesses. Due to their life-threatening nature, CVDs are responsible for a large portion of global fatalities, emphasizing the importance of early medical intervention. Electrocardiograms (ECGs) are essential tools for identifying a wide range of cardiac disorders, and substantial studies have explored automated anomaly detection to aid in preventive care. This research introduces a computational model designed to evaluate ECG signals for predicting heart diseases. The objective is to enhance patient care while simultaneously lowering healthcare expenditures, especially given the increasing costs of medical services and insurance. Utilizing a publicly available dataset containing ECG images of cardiac patients, this study applies a deep learning approach—specifically, the MobileNet framework—to classify four major heart conditions: irregular heartbeat, myocardial infarction, previous myocardial infarction, and normal cardiac activity. The proposed model attained a training accuracy of 97.34% and a validation accuracy of 91.00%, reflecting robust classification capability. Beyond this, the model can also act as a feature extractor to support traditional machine learning algorithms. By addressing the shortcomings of manual diagnostic techniques, the MobileNet-powered system provides an effective and dependable tool to assist medical practitioners in identifying cardiovascular