Classification of ECG patterns using fuzzy rules derived from ID3-induced decision trees

In cardiology, determining whether an electrocardiogram (ECG) is normal or not is sometimes referred to as ECG classification. ECG is the most frequently-used means of cardiac diagnosis. It is the cheapest and the most widely-available; it is also crucial for detecting rhythmic problems. In this paper we derive fuzzy rules for ECG classification from ID3-induced decision trees. The system of fuzzy rules is designed based on 106 ECGs, and it is evaluated using a validation set of 48 ECGs carefully selected by cardiologists. Using the same 106 ECGs for design and the same 48 ECGs for validation, an ID3-generated decision tree yields 73% correct classifications, and a neural network trained with the feedforward cascade-correlation algorithm produces 85.4% correct classifications. On the other hand, the derived fuzzy rules, combined with an optimized defuzzification using the cascade correlation neural network, produces 100% correct classifications.