Abstract

Heart auscultation has served as an important primary measure to identify the pathological condition of cardiovascular system for a long time. However, clinical capabilities to diagnose cardiovascular diseases (CVD) using heart sound (HS) signal have regressed recently. Moreover, the conventional computational approaches have limited progression due to the formidable challenges of HS component segmentation and feature extraction. Thus, a new approach of HS analysis using descriptive string complexity and similarity analysis technique is proposed.

The proposed methodology of analysing HS signal for CVD diagnosis employs key technologies from information theory. These include the coding and encoding descriptor to convert the HS signal to a bio-sequence; parsing based estimators—LZ production complexity and Titchener’s T-complexity and its derivatives T-information and T-entropy to measure the information content of the bio-sequence. Subsequently, similarity metric and conditional entropy are employed to analyze HS signal. Finally the inference networks are proposed to design the CVD diagnosis model. A brief overview of the methodology and the relevant testing results using HS samples from patients will be presented.