Shannon entropy Morlet wavelet Transform (SEMWT) and Kernel Weight Convolutional Neural Network (KWCNN) classifier for arrhythmia in electrocardiogram recordings

Abstract

The automatic detection and classification of life-threatening arrhythmia is life-threatening in treatment of a variety of cardiac diseases. Cardiac arrhythmias are irregular heartbeats that are either too fast (tachycardia) or too slow (bradycardia). Minor alteration in the morphology or dynamics of the Electrocardiogram (ECG) can induce severe arrhythmia events, that can decrease the heart's ability to pump blood and cause breathing difficulties, chest pain, tiredness, and loss of consciousness. A unique deep learning technique for classification of distinct types of arrhythmia utilizing feature extraction is provided in this research. To acquire morphological features, the Shannon Entropy Morlet Wavelet Transform (SEMWT) is done to every heart beat. This work uses Empirical Mode Decompositions (EMDs) with Fuzzy Weight Beetle Swarm Optimization (FWBSO) is introduced for signal noise removal. Then, utilising Kullback-Leibler Divergence Kernel Principal Component Analysis (KLDKPCA) and Dynamic Time Wrapping ECG segments are selected, whereas morphological features from P-QRS-T waves are extracted using SEMWT. SEMWT improves the time and frequency resolution of an ECG signal, making it easier to decode critical information. ECG signal was divided into low-frequency approximation and high-frequency detail components after applying SEMWT. Then, with high accuracy, a Kernel Weight Convolutional Neural Network (KWCNN)-based automated arrhythmia classification is constructed. This work’s resulted are evaluated with performance metrics of Sensitivity (SEN), F-measure, Positive Predictivity (PP) and Accuracy (ACC). Over the whole MIT-BIH Arrhythmias Database, the suggested approach was tested. The suggested classification approach was first tested in MATLAB, with the results compared to those of other methods.