DEEP LEARNING MODEL WITH HYBRID CHAOTIC SAND CAT SWARM OPTIMIZATION AND SAILFISH OPTIMIZER FOR ARRHYTHMIA ECG SIGNAL CLASSIFICATION

Abstract

Arrhythmia of the heart is a collection of erratic heartbeats that result from an anomaly with the coronary electrical structure. The electrocardiogram or ECG measurements can detect cardiovascular rhythmic abnormalities, known as arrhythmias. On the other hand, seeking experts to assess an enormous amount of ECG data consumes an undue amount of medical infrastructure. Deep learning, also known as DL, maybe the better choice for fast and automatic categorization with proper training. The current study presents an innovative hybrid optimizer for the feature selection process in arrhythmias of the heart. The noisy signal of ECG extracted from the dataset of MIT-BIH Arrhythmia was pre-processed by the utilization of finite impulsive response (FIR) and Infinite impulsive response (IIR) to remove the noises of technical as well as biological, leading to misclassification of arrhythmia. The pre-processed wave has redundant features which are removed with the acquisition of the proposed hybrid model optimizer of Chaotic sand cat swarm optimizer with sailfish optimizer. The non-redundant features accumulated from the proposed model are employed under the classifier of the Interpretable Temporal Attention Network (ITA Net). It is further classified into five classes supraventricular ectopic (S), non-ectopic (N), fusion (F), ventricular ectopic (V), and unknown (Q). We can confirm that the proposed approach encouragingly outperforms the current arrhythmia categorization.