ECG based detection of left ventricular hypertrophy using higher order statistics

Abstract

Electrocardiogram (ECG) is a popular non-invasive test record, which shows the electrical activities of the heart. In this paper we propose a novel method to detect left ventricular hypertrophy (LVH) with the use of ECG. Left ventricular hypertrophy is defined as the enlargement of the left ventricle, which is a common disease among hypertension patients. Proposed algorithm uses discrete wavelet transform (DWT) to extract morphological features of ECG signal and exploits higher order statistic (HOS) features including kurtosis, skewness and 5th moment. These features are fed to a support vector machine (SVM) classifier with kernel function of radial basis function (RBF). Our method has been tested on a large database of ECG signals and we have obtained the highest accuracy of 99.6% and sensitivity of 99.4%.