Arrhythmia detection using multi-lead ECG spectra and Complex Support Vector Machine Classifiers

Abstract

Electrocardiograms (ECG) are extensively used for the diagnosis of cardiac arrhythmias. This work investigates the use of machine learning classification algorithms for ECG analysis and arrhythmia detection. Four beat types, Normal (N), Premature Ventricular Contraction (PVC), Atrial Premature Contraction (APC) and Right Bundle Branch Block Beat (RBBB) are simultaneously presented to a Complex Support Vector Machine (CSVM) classifier. The ECG signals are obtained from the St Petersburg INCART 12-lead Arrhythmia Database (INCARTDB). The detection of ECG Wave (P, QRS, T) is performed with the Wave Form Database (WFDB) Software Package which is used to read the annotation files and find the R (peak) location. For feature extraction, the Discrete Fourier Transform (DFT) is used. Fifty Fourier coefficients were selected for reconstructing individual ECG beats. This ensures moderate dimensionality reduction and de-noising of the input vector to the classifier. ECG beats classification is performed using CSVM with several training and test datasets. Sequential Minimal Optimization (SMO) is used to train the CSVM and compute the hyperplane parameters associated with both the real and complex hyperplanes. Cross validation is used for finding the best parameter values of the SVM and the two Gaussian RBK kernel functions. The aim of the study is to establish the advantage of CSVM over standard SVM in simultaneously detecting different types of arrhythmias on the basis of multi-lead recordings following signal compression in the Fourier domain. Implementation of the algorithms was performed in MATLAB. The CSVM classification algorithm provided better performance than the standard SVM classifier. The classification accuracy of the proposed scheme is 98.25% using CSVM. Future work will concentrate on the further development of ECG signal pre-processing using adaptive wavelet algorithms as well as classification with Clifford SVMs.