Hyperellipsoidal classifier for beat classification in ECG signals

Abstract

In this chapter, we study an ECG signal processing and beat classification problem in particular, considering a technique coming with high interpretation capabilities, namely a family of hyperbox classifiers. Different techniques based on a combination of fuzzy clustering and genetic algorithms have been applied to support learning processes in the specific pattern recognition task. Hyperbox classifiers have been investigated for the detection and classification of different types of heartbeats in electrocardiograms (ECGs), which are of major importance in the diagnosis of cardiac dysfunctions. In particular, the learning capacity and the classification ability for normal beats (N) and premature ventricular contractions (PVC) have been tested, with focus on the interpretability aspect. The MIT-BIH arrhythmia database has been used for testing and validating of the proposed method. A total of 26 morphology features have been extracted from ECG and reconstructed VCG signals. Three learning procedures have been developed and analyzed. We have elaborated on several ways of combining fuzzy clustering and genetic algorithm in identifying the optimal hyperboxes and forming a family of hyperellipsoids. The results showed that even a limited number of hyperboxes can increase the geometrical interpretability without a significant reduction of the accuracy of such classifier.