A Novel Hybrid Model to Detect and classify Arrhythmia Using ECG and Bio-Signals

Abstract

In general, arrhythmias, also called cardiac arrhythmia, heart arrhythmia, or dysrhythmias, are abnormal heartbeats which include too fast or too slow. The Cardiovascular Disease (CVD) is a significant cause of death, and the death rate is increasing every year. The Electrocardiogram (ECG) majorly contributes to the CVD diagnosis, providing information about the heartbeat. An automatic detection and classification of arrhythmia performs a significant role in managing and curing cardiovascular diseases. Deep Learning (DL)-based algorithms have emerged as effective solutions in medical applications, particularly in cardiac arrhythmia diagnosis. In this research, a DL-based multi-modal approach is proposed for the classification of cardiac arrhythmia. The MIT-BIH dataset is utilized to evaluate the performance of the proposed method. The proposed method considers physiological signals along with the MIT-BIH dataset to improve accuracy. The Discrete Wavelet Transform (DWT) is used for pre-processing the MIT-BIH dataset. The DL methods of Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) are utilized for classifying cardiac arrhythmia. The proposed method is evaluated using various performance metrics such as Accuracy, Specificity, Sensitivity, F1-score, and Cohen’s kappa.