A model-based approach for arrhythmia detection and classification

Abstract

Automatic real-time ECG patterns detection and classification has great importance in early diagnosis and treatment of life-threatening cardiac arrhythmia [7]. In this paper, we developed an algorithm which could classify abnormal heartbeat at more than 85% accuracy. The ECG data of this research are provided by MIT-BIH Arrhythmia Database from Physionet. We extracted seven features from each ECG record to represent the ECG signal. Furthermore, Support Vector Machine and Multi-Layer Perceptron Neural Network are used for classification. We were able to achieve over 85% accuracy and with only 10% difference between sensitivity and specificity.