FMCW radar accurately detects OLST transitions and captures wobble signatures linked to neuromuscular control and fall risk, establishing proof-of-concept for non-contact radar assessment of balance in older adults.
Key Findings
Results
A CNN-LSTM-Attention model trained on radar-derived range-Doppler maps detected Foot-Up events with 94.0% accuracy and Foot-Down events with 90.7% accuracy.
Performance was evaluated against force plate and motion capture ground truth in a cross-sectional study of 32 healthy adults.
Accuracy improved to 96.9% for Foot-Up and 94.5% for Foot-Down events after logical post-processing.
The model classified OLST phases from synchronized radar, force plate, and motion capture data collected during short (4 s) and long (20 s) OLST trials.
Results
Radar-derived Doppler wobble metrics during the stability phase strongly correlated with OLST duration in long trials.
R2=0.58, p<0.001 for radar-derived Doppler wobble metrics vs. OLST duration.
This aligned with similar trends in force plate metrics (COP ellipse area, R2=0.72, p<0.001) and MOCAP metrics (trunk pitch-period IQR, R2=0.65, p<0.001).
Analysis was conducted on long (20 s) OLST trials.
Results
In short trials, radar-derived features significantly separated the three study-defined cohorts with large effect sizes.
Significant pairwise differences were observed (p<0.05) across cohorts.
Effect sizes ranged from d=1.13 to d=2.72, classified as large.
The study included 15 young participants (≤32 years) and 17 older participants (≥64 years), forming multiple cohorts.
Analysis was conducted on short (4 s) OLST trials.
Methods
The study used a cross-sectional design with 32 healthy adults and collected synchronized radar, force plate, and motion capture data.
The sample comprised 15 young adults (≤32 years) and 17 older adults (≥64 years).
Data were collected during both short (4 s) and long (20 s) OLST trials.
All de-identified data and visualization code are publicly available on PhysioNet.
Force plates and motion capture served as gold-standard simultaneous comparators.
Background
Established clinical methods such as the One-Legged Stand Test are typically performed only annually, missing changes in fall risk between visits.
Current methods rely on supervised, contact- or video-based systems.
Falls are identified as a leading cause of morbidity and mortality in older adults.
The limitation of infrequent assessment motivated evaluation of a privacy-preserving FMCW radar system for autonomous, longitudinal monitoring.
Copeland D, Zhang X, Linton E, Mori B, Namburi P, Anthony B. (2026). Non-contact radar assessment of One-Legged Stand Test for fall risk in aging.. Gait & posture. https://doi.org/10.1016/j.gaitpost.2026.110108