A recognition of ECG arrhytihemias using artificial neural networks

In this study, Artificial Neural Networks (ANN) has been used to classify the ECG arrhythmias. Types of arrhythmias chosen from MIT-BIH ECG database to train ANN include normal sinus rhythm, sinus bradycardia, ventricular tachycardia, sinus arrhythmia, atrial premature contraction, paced beat, right bundle branch block, left bundle branch block, atrial fibrillation, and atrial flutter. The different. structures of ANN have been trained by arrhythmia separately and also by mixing these 10 different arrhythmias. The most appropriate ANN structure is used for each class to test patients´ records. The ECG records of 17 patients whose average age is 38.6 were made in the Cardiology Department, Faculty of Medicine at Selcuk University. Forty-two different test patterns were extracted from these records. These patterns were tested with the most appropriate ANN structures of single classification case and mixed classification cases. The average error of single classifications was found to be 4.3% and the average error of mixed classification 2.2%.