80N as the Optimal Assistive Threshold for Wearable Exoskeleton-Mediated Gait Rehabilitation in Parkinson's Disease: A Prospective Biomarker Validation Study

Abstract

Background and Objectives: Robotic exoskeletons show potential in PD gait rehabilitation. But the optimal assistive force and its equivalence to clinical gold standard assessments are unclear. This study aims to explore the clinical equivalence of the lower limb exoskeleton in evaluating PD patients' gait disorders and find the best assistive force for clinical use. Methods: In this prospective controlled trial, 60 PD patients (Hoehn and Yahr stages 2-4) and 60 age-matched controls underwent quantitative gait analysis using a portable exoskeleton (Relink-ANK-1BM) at four assistive force levels (0 N, 40 N, 80 N, 120 N). Data from 57 patients and 57 controls were analyzed with GraphPad Prism 10. Different statistical tests were used based on data distribution. Results: ROC analysis showed that exoskeleton-measured velocity had the strongest power to distinguish PD patients from controls (AUC = 0.9198, p < 0.001). Other parameters also had high reliability and validity. There was a strong positive correlation between UPDRS-III lower extremity sub-score changes and gait velocity changes in PD patients (r = 0.8564, p < 0.001). The 80 N assistive force led to the best gait rehabilitation, with a 58% increase in gait velocity compared to unassisted walking (p < 0.001). Conclusions: 80 N is the optimal assistive threshold for PD gait rehabilitation. The wearable lower limb exoskeleton can be an objective alternative biomarker to UPDRS-III, enabling personalized home-based rehabilitation.

Keywords: Parkinson’s disease; gait disturbance; gait monitoring; lower limb exoskeletons; wearable device.

Figure 1

(A) Schematic diagram of the lower limb exoskeleton robot, featuring adjustable assistive devices on both feet. (B) Gait parameters detected during patient walking, illustrating the variations in gait parameters throughout the gait cycle. (C) The torque applied to the ankle joint by the exoskeleton in this study is generated by Bowden cables and an ankle joint spring. The Bowden cables apply force at two anchor points on the leg brace, with the ankle joint as the pivot point. The moment arm length is constant at 0.1 m, and its variation during the entire plantarflexion rotation is negligible. Additionally, the tensile force from the Bowden cables must overcome the elastic force of the ankle joint spring. The torque applied to the ankle joint can be expressed as: ${\tau }_a=f·R-{\tau }_s$, where $f$ = tensile force of the Bowden cable (measured in real time by tension sensors at the leg brace anchor points); R = moment arm length (constant at 0.1 m); and ${\tau }_s$ = torque of the ankle joint spring, approximated by ${\tau }_s={\tau }_0+k·{\theta }_a$, where ${\tau }_0$ = initial torque (approximately −2.033 N·m), k = torsional stiffness (approximately 0.033 N·m/°), and ${\theta }_a$= ankle plantarflexion angle.

Figure 2

Sixty patients were enrolled in the PD experimental group and sixty in the healthy control group (three patients were excluded from each group due to not meeting the inclusion criteria).

Figure 3

(A) Area under the curve (AUC) for different gait parameters: The AUCs for various gait parameters (velocity, cadence, left and right strides, and left and right stance phases) are presented. A larger AUC indicates higher reliability and validity. As shown in the figure, velocity has the largest AUC, indicating it possesses the highest reliability and validity among the assessed parameters. (B) Scatter plots of gait parameter comparison between experimental and control groups: The average values of the gait parameters for the experimental and control groups are summarized above. The symbols indicate the statistical significance of differences between the groups: *** represents p < 0.001, and **** represents p < 0.0001.

Figure 4

Correlation between velocity changes and UPDRS-III lower limb score changes: The analysis of the correlation between changes in velocity and differences in UPDRS III lower limb gait sub-scores was conducted using data from the lower limb exoskeleton robot. The correlation coefficient was found to be r = 0.8564 (p < 0.001), indicating a strong relationship. The accompanying figure illustrates this correlation, while the relevant sub-scores from UPDRS III include items 3d, 3e, 5d, 5e, 9a, and 9b, as detailed in the appendix. These findings suggested that monitoring with the lower limb exoskeleton can effectively substitute for UPDRS-III lower limb gait sub-scores, demonstrating its utility in assessing gait characteristics in patients with Parkinson’s disease.

Figure 5

(A). Line graphs of velocity, cadence, stride, and stance phase ratio under different assistance levels. The line graphs display the changes in velocity, cadence, stride, and stance phase ratio for the 57 patients in the experimental group at various assistance levels. This visualization illustrates the trends and patterns in gait parameters as external assistance is varied. (B). Box plots of gait parameters under different assistance levels. The box plots provide a summary of the gait parameters for the experimental group at different assistance levels. Each plot displays the median, quartiles, and potential outliers, offering a clear view of the distribution and variability of the data for velocity, stride, and stance phase percentages. (C). The 95% confidence intervals for gait parameters under different assistance levels. The figure shows the 95% confidence intervals for the gait parameters of the experimental group across different assistance levels. These intervals indicate the range within which we can be 95% confident that the true mean of each parameter lies, highlighting the reliability of the observed differences. In the context of statistical significance: none of * indicates p > 0.05 (ns: not significant), * indicates p < 0.05 (significant), ** indicates p < 0.01 (highly significant), *** indicates p < 0.001 (very highly significant).