Nonlinear Assessment of Gait Signal Complexity in Neurodegenerative Disorders

The human gait cycle undergoes discernible alterations upon the onset of neurodegenerative diseases (NDD) such as Parkinson's, Huntington's, and Amyotrophic lateral sclerosis. Each specific neurodegenerative disorder imparts a distinct influence on human gait dynamics, and precise quantification of these changes holds the potential for accurate methods of NDD detection.Nonlinear entropy algorithms, such as sample entropy (SampEn), find widespread use in physiological signal analysis. SampEn gauges signal complexity by identifying pattern matches within windowed sub-segments of the signal. However, traditional SampEn is notably dependent on user-defined parameters, particularly the tolerance parameter r, leading to inaccuracies in complexity information.SampEn profiling emerges as an alternative concept, eliminating the need for an input r parameter. This data-driven algorithm autonomously generates a set of 'r' values based on the signal's dynamics, yielding a comprehensive SampEn profile. The SampEn profile, containing extensive information about the signal's complexity, serves as a valuable resource for extracting secondary entropy features.In this study, we have contrasted the efficacy of traditional SampEn with SampEn profile-based secondary features such as Total SampEn (TSE) and Median SampEn (MSE), in identifying neurological states. Our findings consistently reveal that secondary features derived from the reduced-parametric SampEn profiling method outperform the traditional parametric SE in distinguishing control cohorts from specific Neurodegenerative Disease (NDD) cohorts.