COMPUTER AIDED DIAGNOSIS OF VENTRICULAR ARRHYTHMIAS FROM ELECTROCARDIOGRAM LEAD II SIGNALS

In this work, we use computer aided diagnosis (CADx) to extract features from ECG signals and detect different types of cardiac ventricular arrhythmias including Ventricular Tachycardia (VT),Ventricular Fibrillation (VF), Ventricular Couplet (VC), and Ventricular Bigeminy (VB).Our methodology is unique in computing features of lower and higher order statistical parameters from six different data domains: time domain, Fourier domain, and four Wavelet domains (Daubechies, Coiflet, Symlet, and Meyer). These features proved to give superior classification performance, in general, regardless of the type of classifier used as compared with previous studies. However, Support Vector Machine (SVM) and Artificial Neural Network (ANN) classifiers got better performance than other classifiers tried including KNN and Naïve Bayes classifiers. Our unique features enabled classifiers to perform better in comparison with previous studies: for VT, 100% accuracy while best previous work got 95.8%, for VF, 100% accuracy while best previous work got 97.5%, for VC, 100% sensitivity while best previous work got 71.8%, and for VB, 100% sensitivity while best previous work got 84.6%