Arrhythmia Classification Based on Adaptive Refined Composite Multiscale Fluctuation Dispersion Entropy

Abstract

To improve the accuracy of electrocardiography (ECG) signal classification and identify abnormal heart rhythms, an arrhythmia classification algorithm based on adaptive refined composite multiscale fluctuation dispersion entropy (ARCMFDE) is proposed. First, an improved QRS complex detection algorithm named the improved Pan-Tompkins algorithm (IPTA) is used. The QRS wave is detected, and the waveform is further processed; then, the signal is decomposed into multiple modal components using variational mode decomposition with the optimized number of decomposition layers (K). Subsequently, the RCMFDE is extracted from the different modal components as a classification feature. Finally, differential evolution (DE) and grey wolf optimization (GWO) are combined to form the hybrid differential evolution-grey wolf pack optimization (DE-GWO) algorithm to optimize the penalty factor c and the kernel function parameter g of the support vector machine for performing pattern recognition. Experimental results show that compared with other methods such as variational mode decomposition (VMD), fluctuation dispersion entropy (FDE), genetic algorithms (GA), and support vector machine (SVM). The proposed classification model has superior performance, with an average accuracy of 96.1%, a sensitivity of 95.9%, and a specificity of 98.7% for four types of heart rhythm recognition. Thus, accurate classification of ECG signals can be achieved using the proposed ARCMFDE-based DE-GWO method.