Integrating Variational Mode Decomposition with Particle Swarm Optimization for feature selection achieved a highest classification accuracy of 97.42% for arrhythmia detection in ECG signals.
Integrating particle swarm optimization for feature selection after variational mode decomposition yields high accuracy (97.42%) for automated ECG arrhythmia classification.
Arrhythmia detection continues to pose significant challenges because ECG signals are non-stationary and susceptible to noise. Signal decomposition, particularly variational mode decomposition (VMD) is advantageous for achieving frequency separation with minimal distortion and is often integrated with metaheuristic optimization to address diverse objectives. However, previous studies have only focused on a single integration workflow, necessitating a comparison of which integration method is more effective. This study proposes a comparison of VMD–PSO–integrated arrhythmia classification methods to assess how integration stages affect feature-extraction quality and ECG classification accuracy. The system was validated using an ECG dataset from the MIT-BIH Arrhythmia Database, which contains 40 patient recordings with five classes. This study focuses on two experiments: PSO for feature selection from VMD decomposition results and optimizing VMD parameters. The results show that differences in VMD parameters and the number of features used in both experiments affect classification quality. The highest accuracy, 97.42%, was achieved in experiment 1 where VMD used fixed parameters (K=5,α=adaptive) and the extracted features were selected using PSO to yield 39 out of 84 features. Building on these findings, the direct comparison of two PSO-based integration strategies provides a novel analysis of their effects on ECG arrhythmia classification, contrasting with prior work restricted to a single optimization aspect.
Mazaya et al. (Thu,) conducted a other in Arrhythmia (n=40). Variational Mode Decomposition (VMD) with Particle Swarm Optimization (PSO) for feature selection vs. VMD with PSO for parameter optimization was evaluated on Classification accuracy. Integrating Variational Mode Decomposition with Particle Swarm Optimization for feature selection achieved a highest classification accuracy of 97.42% for arrhythmia detection in ECG signals.