Investigating the Nonlinear Dynamics of Human Balance Using Topological Data Analysis

Abstract

Understanding the mechanisms behind human balance has been a subject of interest as various postural instabilities have been linked to neuromuscular diseases (e.g., Parkinson's, multiple sclerosis, and concussion). This paper presents a method to characterize an individual's postural stability and estimate of their neuromuscular feedback control parameters. The method uses a generated topological mapping between a subject's experimental data and a dataset consisting of time-series realizations generated using an inverted pendulum mathematical model of upright balance. The performance of the method is quantified using a set of validation time-series realizations with known stability and neuromuscular control parameters. The method was found to have an overall sensitivity of 85.1% and a specificity of 91.9%. Furthermore, the method was most accurate when identifying limit cycle oscillations (LCOs) with a sensitivity of 91.1% and a specificity of 97.6%. Such a method has the capability of classifying an individual's stability and revealing possible neuromuscular impairment related to balance control, ultimately providing useful information to clinicians for diagnostic and rehabilitation purposes.