Main Article Content
Arrhythmias are abnormal rhythms of heart. An irreparable damage is caused due to arrhythmias like Ventricular Tachycardia (VT) and Ventricular Fibrillation (VF). Timely detection of such fatal arrhythmias is vital for the successive application of defibrillation therapies. Heart diseases are diagnosed using different types of tests. Electrocardiography (ECG) testing is one of most efficient and non-invasive method among all. The defined algorithms used for detection and classification of arrhythmias are based on surface Electrocardiogram analysis. However, in this research, an algorithm based on shape statistical and spectral kurtosis features is presented. In this algorithm, shape statistical features along with spectral kurtosis features are applied to the classifier to classify and detect the type of arrhythmia. Various performance measure like accuracy, execution
time and comparison of various performance functions like MSE, SSE, SAE, MAE along with training functions like Trainrp, Trainbr, Trainscg, Trainlm is done in order to identify the most excellent neural network models. The results shows an average accuracy of 90.57% with Feed Forward neural network (FFNN) as the most excellent neural network (NN) for classification of cardiac arrhythmia. The ultimate objective of this algorithm is to contribute as a coherent heart diagnostic and arrhythmia detection tool.
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.