Deep Learning Framework For Early Identification Of Cardiac Abnormalities From Ecg Images

Abstract

Over the past few decades, cardiovascular illnesses have become the leading cause of death worldwide in both industrialized and developing nations. The mortality rate can be decreased through early identification of heart disorders and ongoing clinical monitoring by professionals. Cardiovascular disease is extremely lethal by nature and kills a disproportionately large number of people worldwide. An effective early detection strategy is crucial to preventing cardiovascular disease deaths. An electrocardiogram (ECG) is a vital tool for understanding a variety of human heart problems. This subject has been the subject of numerous investigations to identify heart anomalies for prevention. In order to forecast cardiovascular disorders, this research intends to construct an algorithmic model to analyses ECG tracings. This effort directly affects saving lives and enhancing healthcare at a lower cost. Saving lives and enhancing medical treatment are two immediate effects of this work as health care and health insurance expenses rise globally. In this study, the public ECG picture dataset of cardiac patients was used to harness the potential of deep learning techniques to predict the four main cardiac abnormalities: abnormal heartbeat, myocardial infarction, history of myocardial infarction, and normal person classes. We used the Mobile Net Architecture to build the system for this project, and we were successful in achieving training accuracy of 97.34% and validation accuracy of 91.00%. As a result, the proposed Mobile Net Architecture model classifies cardiovascular diseases with impressive accuracy and can also be utilized to extract features for conventional machine learning classifiers. Bypassing the manual method that produces unreliable and time-consuming findings, the suggested Mobile Net Architecture model can be utilized as a tool to assist physicians in the medical profession in detecting heart disorders using ECG images.