Predicting heart attack using Artificial Neural Network methods

In this thesis, the design and manufacture of an auxiliary decision support system was carried out by using the Multilayer Artificial Neural Network (ML-ANN) model and Electrocardiogram (ECG) signals, which is one of the machine learning methods in the literature, in order to detect the heart attack status of the person. ECG data obtained from the "PTB Diagnostic ECG" data set was used in the study. Firstly, using Pan-Tompkin's algorithm on these ECG signals, noises were removed and different features were obtained. With these obtained features and determined parameters, a ML-ANN model was designed and trained. A separate control circuit has been prepared in addition to the unit in which the trained ML-ANN model is located. With this control circuit, ECG signals were instantly received from the person and these signals were transferred to the unit where the ML-ANN model was located via bluetooth. After obtaining an estimation information about the health status of the person from the transmitted signals with ML-ANN, it is sent to the control card via bluetooth and the health status of the person is presented to the user on the screen. The overall success of the system was 0.86, 0.88, 0.70, 0.96, 0.41, 0.92 and 0.52 for accuracy, positive recall, negative recall, positive sensitivity, negative sensitivity and F1 score, respectively. In line with these results, it has been proven that the system is successful in estimating the detection of heart attack, and another current study on the subject has been brought to the literature.