Abstract

A bounded random process (BRP) model was developed in this paper to interpret why approximate entropy (ApEn) can not describe the stochastic characteristic of Brownian motion time series correctly. Low ApEn value, generally implying presence of determinacy, also existed in Brownian motion series, a stochastic process. The BRP model investigated this phenomenon through quantifying the relationship between ApEn and a parameter of BRP model. BRP model was then applied to analyze electrocardiograph (ECG) time series from 60 healthy subjects and 60 myocardial infarction (MI) patients. ApEn of the healthy group had a close relationship with the parameter of BRP model, while this relationship could not be found in MI patient group. Grounded on combination of BRP and ApEn, a classifier was designed to assist to diagnose the MI patients. ROC curve and classification figures verified the classifier.

Key words: Bounded random process, approximate entropy, Brownian motion, heart rate variability.