Multiresolution inter-sample and inter-lead eigen error features for classification of cardiac diseases

This paper presents a new technique to estimate diagnostic features in multilead electrocardiogram (MECG) signal. The technique uses the eigen analysis of the significant sub-band matrices of analyzed MECG and synthetic MECG for evaluation of features. The sub-band matrices are obtained using the multiresolution analysis of MECG. For each of the significant sub-band matrix, the inter-sample eigen error (ISEE), the inter-lead eigen error (ILEE) and the singular value error (SVE) features are evaluated. The multilayer perceptron (MLP) neural network and the support vector machine (SVM) classifiers are employed to classify the MECG features into three cardiac diseases (heart muscle disease (HMD), myocardial infarction (MI) and bundle branch block (BBB)) and healthy control. The result reveals that, for MI, HMD and BBB, the proposed technique has better performance with sensitivity values of 99.43%, 99.77% and 97.78%, respectively.