Performance Analysis of Arrhythmia Detection using Multiclass Classifier

Abstract: In this paper, the most authentic and efficacious method for cardiac arrhythmia classification using Multiclass Support Vector Machine (MSVM) is presented. The authors have considered classification of 6 beat types such as normal sinus rhythm (N), Premature Ventricular Contraction (PVC), Right Bundle Branch Block (RBBB), Left Bundle Branch Block (LBBB), Tachycardia (TA) and Bradycardia (BR) by implementing MSVM classifier. Radial Basis Function (RBF) kernel with 5 fold cross validation and zero offset value is used for adjusting kernel values. A total of 24 ECG records are used to collect different types of beats. To feed the classifier the features adopted where QRS complex, RR interval, R amplitude, S amplitude and T amplitude. The MSVM classifier performance is measured in terms of accuracy, sensitivity and specificity. The classifier demonstrates its effectiveness and is found to be highly accurate in ECG classification.

Keywords: Accuracy, cardiac irregularity, classifier, ECG, MSVM, sensitivity and specificity