2026 年 21 巻 2 号 p. 26-00068
Traditional stopwatch-based assessments such as the Timed Up and Go (TUG) test may lack sensitivity for high-functioning older adults because of ceiling effects. This study evaluated the potential of a markerless motion capture (MMC) framework using a single monocular camera to quantify subtle trunk kinematics during TUG and to explore motor control deficits that may be imperceptible to human observation. Thirty-four community-dwelling older adults (17 fallers and 17 non-fallers) performed TUG under a maximal-effort condition to impose biomechanical stress. MediaPipe-based pose estimation was used to extract the trunk center-of-mass (COM) trajectory, and a systematic screening of 18 metrics spanning temporal, variability, and smoothness domains was conducted to identify candidate biomarkers. Total time did not distinguish between groups, whereas the Log Cumulative Jerk (LCJ) in the return walk phase showed the largest effect size (Cohen's d = 0.68), although the between-group difference was not statistically significant under non-parametric testing (Mann-Whitney U, p = 0.130). In this context, lower LCJ may reflect a more constrained movement pattern consistent with a possible “stiffening strategy,” rather than improved coordination. LCJ yielded the highest classification performance among the tested parameters (AUC = 0.65), indicating modest relative discrimination in this sample. These exploratory findings suggest that markerless assessment of movement smoothness may provide complementary information beyond conventional time-based measures and may support hypothesis-generating, non-contact fall-risk screening in community settings.
