Clinical Entropy based PCA for Multi-lead Electrocardiogram Signals

Abstract

In this work, an information theoretic approach is proposed for principal component analysis (PCA) of multi-lead electrocardiogram signals. Clinical information is evaluated from the inverse of the diagonal eigenvalue matrix. It is termed as Clinical Entropy (Centropy). Clinical entropy (Centropy) based PCA method shows improved performance compared to the conventional PCA. The proposed method exhibits superior signal quality with higher cross correlation (CC), lower percentage root mean square difference (PRD) and lower root mean square error (RMSE) values. Keywords—ECG, PRD, PCA, Entropy