Adaptive Techniques for Cardiac Arrhythmia Detection by Using Artificial neural Network

Abstract

95 ECG signals were collected from MIT-BIH Arrhythmia database in Physionet bank in Physionet website. 25 of these samples are said to be abnormal and the rest 70 sample are normal sinus rhythm ECG. These samples were processed to remove baseline wonder and high frequency noise using Fast Fourier Transform and Discrete Wavelet Transform. The Matlab program was also developed to extract the main feature of the ECG signal during a set period of time. Starting with segmenting the signal into three smaller samples then detecting the R peak knowing that it is the easiest feature to detect because it’s the highest peak. The rest of the features were collected easily once R peak is known. Artificial Neural Network was used in the classification step. The normal signal was given the number 1 while the abnormal sample was 0 in the construction of ANN. All the features were presented in an excel file and used in the development of the ANN. All these data were used in the training of ANN and the results was collected and discussed.