Diagnosis of Cardiac Disorders in ECG Signal using Artificial Neural Network Algorithm

Abstract

Heart disease was and still one of the critical disease that may affect our normal heart activity. Electrocardiograph (ECG) which represent heart cycle contain many components in form of waves, intervals, segments etc. In years, many techniques were developed to detect heart disorder and in this paper a method of cardiac arrhythmia detection and classification was developed to differentiate between normal an abnormal ECG signals, this process had included three main steps, First as known all signals even it was biological or other need to remove all the noises that may contained in this case filters were designed to remove these noises, and each kind need specific filter to be remove for example band-reject filter with cutoff frequencies (57Hz-43Hz) was designed to remove power line interferences and band-pass filter with cutoff frequencies (0.05Hz-30Hz) was designed to remove patient motion and patient chest movement. In the next step variables were calculated and features were extracted using a technique called wavelet transform algorithm which decompose the signals using many functions called wavelets, the one had been used was daubechies, It used because it has similar formulation. The extracted features and parameters were used as an input to the final stage, which is the artificial neural network (ANN) algorithm which is an algorithm works like the human brain when dealing with specific tasks, also it gains knowledge by training and evolving. The used signals were downloaded from an online website Massachusetts Institute of Technology Beth Israel Hospital (MIT-BIH) arrhythmia database which used to train and test the classifier. Classifier performance was calculated using four statistical values accuracy (98.7), sensitivity (98.7), positive predicatively (98.6), specify (90.91). The results showed new illations and support implemented ideas which will help the researcher in the future to compare between the found results and the current one, so noticing the progress, all this was implemented using Mat Lab (R2014b) program.