This study aims to design a computer program to detect myocardial ischemic heart defects through electrocardiogram (ECG) signal patterns and their accuracy. Myocardial ischemia is a heart disorder caused by the narrowing of the blood vessels in the walls of the heart. The method used is a backpropagation-based artificial neural network (ANN) based on MATLAB/Simulink. The input data is trained to recognize the target pattern of the ECG signal based on the potential and time in the ST segment. The optimal weight of the results of the ANN backpropagation algorithm is used in the process of testing the ECG signal pattern to obtain the ANN output. The ANN output was analyzed for potential depression or elevation to identify normal heart or myocardial ischemia. The results of the training show that from several architectures that have been tested, the optimal ANN architecture is 1 hidden layer with 11 hidden units. These results are obtained from the epoch parameter and the mean square error (MSE) value as well as the accuracy of each architecture. The backpropagation ANN learning process requires 8 epochs to achieve the performance goal with MSE 4.03 × 10-9. The system can recognize target patterns with a training accuracy of 99.82%. The test results of the ANN program identification system can detect myocardial ischemia and normal heart abnormalities with an accuracy of 86.7%. Some data were not detected because the ANN output did not meet the criteria for cardiac ischemia or normal myocardium on the ECG signal. Based on the accuracy of the ANN program identification system, the detection of myocardial ischemia rhythm ECG signal patterns using ANN can be said to work well.
Read full abstract