Detection of Heart Diseases Using Deep Learning Techniques

Abstract

Electrocardiography is an effective tool for detecting heart diseases or predicting heart diseases, and previous researchers have approved it as an effective tool in diagnosis. This early diagnosis's essential benefit is reducing deaths due to heart disease because the heart is the most critical part of the human body. From this Starting point, this paper used electrocardiography to diagnose and predict heart disease. A system that supports deep learning by using Convolutional Neural Network and the use of the most critical global data set approved by previous researchers was proposed to diagnose or predict the four most critical pathological conditions, namely (STT abnormalities, myocardial infarction (MI), arrhythmias, and Conduction disturbances and abnormalities) The proposed system goes through three primary stages (processing, classification, and prediction), where CNN deep learning algorithms of the design the proposed system. The data set was used. PTB-XL for calculating healthy and infected samples for complete system training, testing, and prediction. The proposed system achieved good results with a sensitivity of 72.3%, a specificity of 73.90%, an accuracy of 91.33%, an accuracy of 88.69%, and an f1 score of 92.51%.