Efficient Electrocardiogram-based Arrhythmia Detection Utilizing R-peaks and Machine Learning

Abstract

The rise in heart-related diseases has led to a need for proper automatic diagnosis methods to identify irregular heart problems. It has proven to be challenging to promptly and accurately diagnose many complicated and interferential symptom diseases including arrhythmia. Recently, thanks to the evolution of artificial intelligence (AI) and the advance in signal processing, automated arrhythmia detection has become easier and widely applied for physicians and practitioners with machine learning (ML) techniques and the only use of electrocardiograms (ECG). In this paper, we propose an ECG-based machine learning arrhythmia detection approach that exploits R-peak detection and machine learning. Our proposed solution targeting a binary classification of heartbeats employs an efficient R-peak detection that uses a Butterworth bypass filter, Ensemble Empirical Mode Decomposition (EEMD), and Hilbert Transforms (HT) for processing ECG signals, and applies the most effective machine learning algorithm among typical ML algorithms to improve the performance of the arrhythmia diagnosis. In order to select the most suitable one with the highest achievable performance, typical ML algorithms such as BG, BS, KNN, and RF were investigated. A popular public dataset, MIT-BIH Arrhythmia, is used for the numerical experiments. The attained results prove that our developed solution outperforms the notable traditional algorithms and it offers the best performance with an accuracy of 93.4%, a sensitivity of 95.4%, and an F1-score of 96.3%. The high obtained F1-score implies that our solution can overcome the data imbalance to detect arrhythmia correctly and be effective in practical clinical environments.