Markerless computer vision demonstrated excellent agreement with manual stopwatch measurements.
In clinical practice, mobility and balance assessments are traditionally timed manually using a stopwatch, which can introduce observer error and variability between assessors.
Marker-based motion capture systems can improve objectivity, but often require expensive equipment and specialized settings, limiting clinical accessibility.
Markerless computer vision (MLCV) motion capture offers a low-cost, objective alternative that may improve the consistency and efficiency of clinical assessments.
MLCV also has potential applications in telerehabilitation.
Before clinical implementation, MLCV-derived measurements must demonstrate agreement with conventional assessment methods.
Background
Methods
Results: ICC
PRESENTER
Nabiha Chowdhury
Nabiha Chowdhury, Julia Raphael, Dr. Bok kyu Kim, PhD, PT
Department of Physical Therapy, SUNY Upstate Medical University
Summer Undergraduate Research Fellowship Program (SURF) 2026
These findings support the feasibility of MLCV as an objective method for timing functional mobility assessments.
Secondary analysis of a pilot study testing the effects of taekwondo on balance and mobility
N= 14
11 non-disabled older adults
3 older adults with Parkinson's disease
Data were collected across multiple assessment timepoints, with two trials completed for each functional outcome measure
Participants completed four functional outcome measures
Five Times Sit-to-Stand (5XSTS)
Figure-8 Walk Test (F8WT)
Performance was recorded simultaneously using a high-resolution, high-speed action camera and a manual stopwatch.
Reference Measurements
Manual stopwatch measurements were collected during live testing (LT) and video review (VR).
Markerless Computer Vision Analysis
Video recordings were analyzed using an MLCV algorithm to automatically detect movement events and calculate task completion times.
Data Analysis
Agreement between measurement methods was evaluated using:
Intraclass Correlation Coefficient (ICC)
Bland–Altman analysis
Computer vision (CV) and LT were each compared with VR, which served as the gold standard.
Results: Bland-Altman Analysis
Tap or take a picture for the full poster and paper!