ECG Based Heart Disease Classification: Advancement and Review of Techniques

Abstract

The heart plays a crucial role in pumping blood to our organs. However, without early detection of heart disease (HD), the outcomes can be fatal. That’s why it is crucial to emphasize the significance of prior detection and accurate classification of HD for effective diagnosis. One important diagnostic technique that has emerged is electrocardiography (ECG). Analyzing and processing ECG signals greatly contribute to diagnosing HD. In this comprehensive review, we discuss recent advancements in HD classification using various machine learning, deep learning, and ensemble learning algorithms such as SVM, CNN, XGBoost, etc. We delve into the advantages and disadvantages of each algorithm while providing an extensive overview of state-of-the-art studies that have utilized these algorithms on datasets MIT-BIH and PTB-XL. Among these models, CNN consistently achieved remarkable accuracy across both datasets with an average of 98.83%. The LSTM model excelled in sensitivity on the PTB-XL dataset with an average recall rate of 88.01%. Additionally, SVM and k-NN models delivered competitive accuracy rates of 96.20% and 97.50%, respectively, on the MIT-BIH dataset. We also explore GMM and K-Means clustering techniques to provide insights into their potential applications in classifying HD. It’s worth noting that existing research often focuses on individual methods without comprehensive comparisons between available algorithms. Our paper aims to bridge this gap by offering a concise overview that encompasses various methods’ strengths, limitations, and advancements, providing valuable insights for future studies.