Emergency diagnosis of Myocardial infarction (MI) by artificial neural network

Myocardial infarction is one of the most common diseases with high mortality and morbidity in human beings. Iranian health ministry official statistical analyses show that the most frequent cause of death in the country after accidents is Myocardial infarction (MI). Inappropriately long patient time delay is the main cause for undesirable pre-hospitalization delay. We decided to apply artificial neural network to decrease the prehospitalization phase time. We used clinical and medicinal parameters taken for 267 persons from ekbatan hospital of Hamadan. We carried out Chi square test with SPSS software for 56 parameters. With regard to the results of this analysis we selected 7 parameters that had the lowest sig for ANN analysis (among parameters, whose sig were less than 0.05). Selected parameters of 267 persons were applied for training network with Levenberg-Marquardt Learning Algorithm. Learning rate was 0.1. The training process finished at around 46 epochs; assembling and training of artificial neural network was done by Matlab software r2009a. Best validation performance was9.06 ∗ 10−15. After plotting the ROC curve, the area under ROC curve was measured to estimate the diagnostic performance; area under roc curve for this analysis was 1.We plotted relation between target value and output of trained neural network for training, validation & test dataset statistically; R value for all of them was 1. Physical examination and accurate ECG interpretations, cardiac biomarkers are equally valuable in the initial evaluation of patients with non-traumatic chest pain. Because quick detection of MI is very vital for patient and these evaluation need more time, we decided to apply ANN for quick and reliable detection of MI. Therefore by using trained ANN we can predict MI quantitatively without requirement of much time. Key-Words: myocardial infarction, diagnosis, heart disease, Artificial Neural Network