ECG Arrhythmia Classification Using Convolutional Neural Networks

Abstract

Cardiac arrhythmias occur in a short duration of time which can’t be distinguishable by a human eye. Cardiac arrhythmia detection is a tedious task since slight changes in ECG signal may lead to life-threatening diseases. Diagnosis and medication at an early stage could help to the decrease the high mortality rate among the heart patients. This paper presents an accurate technique for the classification of five types of ECG arrhythmia namely Premature ventricular contraction(V), Normal (N), Left bundle branch block (L), Right bundle branch block (R), Paced (P). This technique incorporates convolutional neural networks (CNN) that combine both feature extraction, classification into a single body which restricts the use of complex feature extraction techniques like DTCWT (Dual tree complex wavelet transform) and a separate classifier to classify these features into appropriate classes. The performance of the proposed technique is assessed by using MIT-BIH arrhythmia database. Average classification accuracy of 95.43% is obtained which is superior to many other algorithms proposed in the literature.