Abstract
Background
Objective assessment of functional mobility is helpful in evaluating disability and treatment outcomes in patients with degenerative spine disease. Traditional marker-based motion capture offers laboratory-grade accuracy but is limited by complex setup, soft-tissue artifact, and practical challenges in clinical settings. Consequently, there is a gap in scalable, accurate, and clinically feasible methods to objectively assess spine-related functional performance in real-world settings. Markerless motion capture, powered by computer vision and deep-learning algorithms, enables marker-free, rapid, and reproducible assessment of three-dimensional kinematics in clinic settings.Purpose
To evaluate the concurrent validity of a markerless motion capture system relative to a widely accepted reference method (marker-based motion capture) during gait and balance tasks in patients with spine pathology.Methods
A total of 116 participants (93 spine patients, 23 healthy controls) performed standardized gait and balance tasks captured simultaneously by marker-based and markerless systems. Agreement was examined using intraclass correlation coefficients (ICC), Bland-Altman analysis, root mean square deviation (RMSD), and Pearson correlation.Results
Across 580 gait and 348 balance trials, markerless motion capture demonstrated good to excellent agreement (ICC > 0.75) for trunk, lumbar, pelvic, and lower-extremity kinematics, with RMSD values ≤ 5°. Dynamic sagittal vertical axis (SVA) exhibited excellent reliability (ICC = 0.90, 95% CI: 0.84-0.94). Bland-Altman analysis showed minimal systematic bias. Balance assessments revealed excellent agreement for head sway in sagittal and coronal planes (ICC ≥ 0.97). However, transverse-plane rotations at the hip, knee, and ankle demonstrated poor agreement (ICC Conclusions
Markerless motion capture provides quantifiable, accurate, and clinically applicable measures of spine-related movement. Its efficiency, scalability, and accuracy support objective evaluation of disability, enable individualized rehabilitation, and facilitate longitudinal outcome monitoring, highlighting its potential for integration into standard clinical workflows and bridging the gap between laboratory-grade biomechanics and routine spine care.