Nonlinear characteristics of gait signals in neurodegenerative diseases

TL;DR

This paper explores gait signal patterns in neurodegenerative diseases like Parkinson's and ALS, using ratios of left and right limb data to better identify disease types.

Contribution

A novel method for analyzing gait signals using ratios of left and right limb data to distinguish between neurodegenerative diseases.

Findings

Ratios of left to right gait sequences correlate strongly with standard deviation and coefficient of variation, aiding disease classification.

Median filtering reduces noise and stabilizes gait data, improving analysis accuracy.

ALS patients show the highest complexity and entropy in gait signals, followed by HD and PD.

Abstract

Based on the asymmetric characteristics of left and right movements in patients with neurodegenerative diseases and their inherent coupling relationships, as well as the inevitable internal connection between them according to the principles of mechanical kinematics, and a processing method for the ratio of gait signals to left and right limb data is proposed. Using gait time series data collected from left and right limbs via pressure-sensitive insoles, a comparison was conducted among patients with Parkinson's disease (PD), Amyotrophic Lateral Sclerosis (ALS), Huntington's disease (HD), and a healthy control group (Ctrl) in terms of the average, standard deviation, and coefficient of variation of the left and right sequences, as well as the ratios between them. It was discovered that there exists a close correlation between the ratios of left to right sequences and the actual standard deviation and coefficient of variation of these sequences. These ratios can be utilized for identifying the categories of PD, ALS, and HD patients. After using a median filter (n = 3) to filter four sets of stride ratio data (Ctr1, A1s, PD, and HD), it was found that the data before filtering generally showed significant fluctuations, with many peaks and valleys, indicating that the original data may contain a lot of noise or outliers. In contrast, the filtered data showed relatively smaller fluctuations and a smoother curve, indicating that the filtering process effectively reduced noise in the data and enhanced its stability. The raw data distribution for the left and right limbs of patients with PD, ALS, HD, and the Ctrl was relatively large, posing certain difficulties in analyzing the patients' diseases. The use of the ratio of left to right data effectively improves the discreteness of the data. The ranking of CO complexity features from highest to lowest is ALS, HD, PD, and Ctrl. The ranking of sample entropy features from largest to smallest is ALS, HD, PD, and Ctrl. The ranking of wavelet coefficient features from largest to smallest is ALS, PD, HD, and Ctrl.

Figures