Development of a Computer-Aided Application for Analyzing ECG Signals and Detection of Cardiac Arrhythmia Using Back Propagation Neural Network - Part I: Model Development

Electrocardiogram (ECG) is a graphic recording of the electrical activity produced by the heart. The accuracy of any electrocardiogram waveform extraction plays a vital role in helping a better diagnosis of any heart related illnesses. We present a computer-aided application model for detection of cardiac arrhythmia in ECG signal, which consists of signal pre-processing and detection of the ECG signal components adapting Pan-Tompkins and Hamilton-Tompkins algorithms; feature extraction from the detected QRS complexes, and classification of the beats extracted from QRS complexes using Back Propagation Neural Network (BPNN). The application model was developed for ECG signal classification under ‘Normal’ or ‘Abnormal’ heartbeats to detect cardiac arrhythmia in the ECG signal. The model was trained with standard arrhythmia database of Massachusetts Institute of Technology Division of Health Science and Technology/Beth Israel Hospital (MIT-BIH), and taking into account the Association for the Advance of Medical Instrumentation (AAMI) standard. The performance of the developed application model for classification of ECG signals was investigated using the MIT-BIH database. The accuracy of detection and extraction of the signal components and features (based only on the MIT-BIH database used) shows that the developed application model can be employed for the detection of heart diseases in patients.