ECG Arrhythmia Classification Using Vague C-Means Clustering Based on Multiresolution Analysis

Abstract

The accurate classification of ECG arrhythmias is crucial for diagnosing heart diseases. The detection and classification of arrhythmia rely on several key factors, including the specialist's experience level, work intensity, and time consumption. These factors are critical determinants of the accuracy and effectiveness of the diagnostic process, which, in turn, directly impacts the patient's health outcomes. Artificial intelligence-based computer-aided diagnosis systems have made great progress in ECG arrhythmia classification in recent years. In this study, ECG arrhythmia classification was performed using a vague c-means clustering algorithm. The data set used was obtained from the MIT-BIH ECG Arrhythmia Database. The data set consisted of 318 patterns which are RR intervals Experiments were performed with different parameters of vague c-means clustering to achieve the highest classification performance. In addition, experiments were repeated using fuzzy c-means clustering for comparison. Furthermore, a frequency-based feature set using multiresolution analysis based on discrete wavelet transform was obtained. An ECG classification task was realized with vague c-means clustering on this frequency-based dataset. The best results were obtained as 87.5%, 80%, and 84% for classification accuracy, sensitivity, and positive predictive value, respectively.