Evaluating the concurrent validity and test-retest reliability of markerless motion capture for static postural...

Markerless motion capture presents potential as a clinically feasible biomechanical assessment tool, although its utility for postural control measurement remains largely unclear. This study assessed the concurrent validity and test-retest reliability of markerless motion capture for measuring static postural control, as well as its robustness to clinically relevant conditions. Fifty-three healthy adults performed three quiet standing tasks across two sessions, spaced ∼one week apart. Tasks were performed under two conditions: standing alone (unaccompanied), and with a person standing nearby simulating clinical supervision (accompanied). Data were simultaneously recorded from eight synchronized video cameras and a force plate, with the Centre of Mass (COM) of the pelvis segment estimated from video using Theia3D and the Gravity Line Projection (GLP) estimated from ground reaction forces. Cross-correlations and Root Mean Square Differences (RMSD) were used to evaluate concurrent validity between the COM and GLP signals. Intraclass and Spearman's correlations were used to assess test-retest reliability and associations across conditions for seven discrete COM parameters. The COM and GLP signals demonstrated strong agreement across all tasks/conditions (mean zero-lag cross-correlation = 0.89-0.99; RMSD < 0.13 cm). Test-retest reliability was generally higher in the unaccompanied condition, particularly for the 95% confidence ellipse area (ICC(A,1) = 0.75-0.89) and mediolateral path length (ICC(A,1) = 0.63-0.84). Most COM parameters were positively correlated across conditions (r = 0.37-0.81). Overall, markerless motion capture produced valid and reliable measures of static postural control. Although decreased reliability in the accompanied condition warrants caution, the strong validity and associations between conditions support future clinical investigations.