Are Strategies Favoring Pattern Matching a Viable Way to Improve Complexity Estimation Based on Sample Entropy?

• Published in: Entropy (Basel, Switzerland) Vol. 22; no. 7; p. 724
• Main Authors: Porta, Alberto, Valencia, José Fernando, Cairo, Beatrice, Bari, Vlasta, De Maria, Beatrice, Gelpi, Francesca, Barbic, Franca, Furlan, Raffaello
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI 30.06.2020; MDPI AG
• Subjects: cardiovascular control; conditional entropy; heart rate variability; information dynamics; systolic blood pressure; time series analysis; time series analysis; autonomic nervous system; heart rate variability; Parkinson disease; conditional entropy; information dynamics; cardiovascular control; systolic blood pressure; head-up tilt
• ISSN: 1099-4300; 1099-4300
• DOI: 10.3390/e22070724

It has been suggested that a viable strategy to improve complexity estimation based on the assessment of pattern similarity is to increase the pattern matching rate without enlarging the series length. We tested this hypothesis over short simulations of nonlinear deterministic and linear stochastic dynamics affected by various noise amounts. Several transformations featuring a different ability to increase the pattern matching rate were tested and compared to the usual strategy adopted in sample entropy (SampEn) computation. The approaches were applied to evaluate the complexity of short-term cardiac and vascular controls from the beat-to-beat variability of heart period (HP) and systolic arterial pressure (SAP) in 12 Parkinson disease patients and 12 age- and gender-matched healthy subjects at supine resting and during head-up tilt. Over simulations, the strategies estimated a larger complexity over nonlinear deterministic signals and a greater regularity over linear stochastic series or deterministic dynamics importantly contaminated by noise. Over short HP and SAP series the techniques did not produce any practical advantage, with an unvaried ability to discriminate groups and experimental conditions compared to the traditional SampEn. Procedures designed to artificially increase the number of matches are of no methodological and practical value when applied to assess complexity indexes.