Cardiac arrhythmia classification using multi-granulation rough set approaches

Cardiovascular disease is a most important reason for human death in modern society. Electrocardiogram (ECG) signal deals with valuable information about functioning of the heart. For that reason, ECG investigation signifies an efficient way to identification and treat different types of cardiac arrhythmia diseases. Nowadays various pattern classification methods has been developed for the classification of ECG signals. These classification methods helps to physician for diseases diagnosis. A multi-granulation rough set (MGRS) has become a new direction of rough set theory, which is based on multiple binary relations on the universe of discourse. In this present study, Multi-granulation rough set based classification approaches (Pessimistic Multi-Granulation Rough Set (PMGRS) and Optimistic Multi-Granulation Rough Set (OMGRS)) are applied to mine appropriate rules to explore better decision making process. The experiments were conducted on the ECG data from the Physionet arrhythmia database to classify five kinds of normal and abnormal signals. In the classification process, Feature extraction played an important role. And we have used two kinds feature extraction methods (1) Pan Tomkins (PT) feature extraction method. This method used to extract the morphological features are P, Q, R, S, T peak intervals, which is also used to determine heart rate. (2) Wavelet transform (WT) feature extraction method. This method used to extract the wavelet coefficients. Both two methods are successfully applied to (ECG signal) classification. The proposed multi-granulation rough set rule based classification methods is validated using the first 24 channel of the ECG signal records of the MIT-BITH arrhythmia database, and achieves finding high accuracies. Experimental results show that the proposed classification techniques significantly outperforms other well-known techniques.