Discrimination of cardiac arrhythmias using a fuzzy rule-based method

The authors propose a relatively simple but robust fuzzy rule-based method to accurately distinguish between normal sinus rhythm (NSR), malignant ventricular fibrillation (VF) and ventricular tachycardia (VT). The method first subjectively identifies the "qualitative" features of pre-processed electrocardiogram (ECG) signals necessary to discriminate between NSR, VF and VT. Second, the parameters of fuzzy sets, namely the real-valued membership functions for "small", "medium" and "large", are defined so that qualitative features are quantitatively interpreted. Finally, a set of fuzzy "if-then" rules is constructed to discriminate between NSR, VF and VT which uses the defined membership functions. This technique was applied to portions of the MIT/BIH Malignant Ventricular Arrhythmia Database and results show correct classification for 94.3% of NSR, 82% of VT and 78% of VF events.<>