Quantifying Static Postural Control Impairment after Stroke using Markerless Motion Capture

Markerless motion capture offers a practical approach to biomechanical assessment in clinical populations outside of controlled laboratory environments. The purpose of this study was to evaluate the utility of markerless motion capture for quantifying static postural control impairment in individuals with stroke. Participants with subacute stroke (n=26) and healthy controls (n=171) performed a 30-second quiet standing task across three data collection sites in Kingston, Ontario, Canada. Video data were processed using Theia3D, and pelvis center of mass (COM) trajectories were extracted. Five discrete COM parameters were computed and transformed into Z-scores using normative models developed from the healthy population. An overall Task Score was calculated by aggregating parameter Z-scores. Forty-six percent of participants with stroke were identified as impaired on the task (Task Score > 95th percentile of the healthy distribution). At the parameter level, impairments were more prevalent in the mediolateral (ML) than anteroposterior (AP) direction (root-mean-square distance: 50% impaired ML vs. 12% impaired AP; mean velocity: 46% impaired ML vs. 35% impaired AP). Task scores and several task parameters demonstrated significant correlations with Berg Balance Scale scores (r = -0.64 to -0.81) and Functional Independence Measure Motor Sub-Scale scores (r = -0.62 to -0.77), with poorer performance associated with lower clinical scores. These findings demonstrate the utility of markerless motion capture for quantifying static postural control impairment following stroke. Future work should extend this framework to evaluate impairment in more dynamic postural control and gait tasks.