Classification of Cardiac Arrhythmias Using Interval Type-2 TSK Fuzzy System

The paper proposes a method to construct type-2 Takagi–Sugeno-Kang (TSK) fuzzy system for electrocardiogram (ECG) arrhythmic classification. The classifier is applied to distinguish normal sinus rhythm (NSR), ventricular fibrillation (VF) and ventricular tachycardia (VT). Two features of ECG signals, the average period and the pulse width, are inputs to the fuzzy classifier. The rule base in the fuzzy system is constructed from training data. We also present the method using fuzzy C-mean clustering algorithm and the back-propagation technique to determine parameters of type-2 TSK fuzzy classifier. The generalized bell primary membership function is used to examine the performance of the classifier with different shapes of membership functions. The results of experiments with data from the MIT-BIH Malignant Ventricular Arrhythmia Database show the classification accuracy of 100% for NSR signals, 93.3% for VF signals, and 92% of VT signals.