100 Section - Transdiagnostic Gait Phenotypes Identified Using a Large-Scale Robotic Gait Measurement System

Wednesday, May 20, 2026

6:00 PM - 7:30 PM PT

Presenting Author/Autor expositor(s)

KS

Kotaro Shimizu, PhD

Research Fellow
Research Institute of National Rehabilitation Center for Persons with Disabilities
Tokorozawa, Saitama, Japan

Co-Author/Coautor(s)

MH

Mikito Hikosaka, PhD

Research Fellow
Research Institute of National Rehabilitation Center for Persons with Disabilities
Tokorozawa, Saitama, Japan
YO

Yuya Ohhashi, MS

Graduate student
Shibaura Institute of Technology
Ohmiya, Saitama, Japan
NK

Noritaka Kawashima, PhD

Section Chief
Research Institute of National Rehabilitation Center for Persons with Disabilities
Tokorozawa, Saitama, Japan

Objectives : Gait disturbance arise in a wide range of neurological and musculoskeletal disorders, yet gait characteristics vary substantially across diagnoses and severity levels. Traditional optical motion-capture systems provide detailed kinematic assessments but are costly, non-portable, and unsuitable for large-scale clinical deployment. As a result, comprehensive datasets capturing gait kinematics across diverse patient populations remain scarce. This study used a novel robotic measurement system equipped with a depth camera and skeletal pose-estimation technology to build a large clinical gait database and investigate cross-diagnostic gait mechanisms using multivariate analysis.

Design: We analyzed 10-m walking from 1,768 individuals, including healthy adults, children, and patients with neurological and musculoskeletal conditions. The robotic system recorded three-dimensional movement trajectories, from which spatiotemporal parameters and joint kinematics were derived. Exploratory factor analysis extracted latent components of gait, which were subsequently used for cluster analysis and structural equation modeling (SEM) with gait speed as the outcome.

Results: Eight factors were identified: (1) knee-dominant limb swing, (2) vertical head displacement, (3) hip-dominant limb swing, (4) foot-generated distal propulsion, (5) asymmetry, (6) variability of joint angles, (7) forward trunk inclination, and (8) lateral limb swing. Clustering revealed subgroups dominated by healthy adults, healthy children, and patient groups reflecting both diagnostic categories and severity levels. SEM showed excellent explanatory power for gait speed (R² = 0.925) with generally acceptable to good model fit (CFI = 0.983, TLI = 0.960, RMSEA = 0.069, SRMR = 0.21).

Conclusion: Factors associated with compensatory strategies—particularly altered head/trunk motion and lateral limb movement—emerged as key descriptors of patient-specific gait patterns. The SEM findings demonstrate that gait speed, a central indicator of gait disability, can be explained by a concise combination of latent components. These results provide a conceptual foundation for simplified clinical gait assessments and next-generation practical gait-measurement devices.