Gait analysis plays a vital role in clinical care by aiding in the diagnosis, treatment planning, and rehabilitation evaluation of movement disorders. Traditional marker-based motion capture systems, however, are hindered by time-consuming setup procedures, high costs, and significant space requirements, limiting their widespread adoption in clinical settings. Recent advances in markerless motion capture technology offer a promising alternative by eliminating the need for markers and extensive setup time. While these systems have demonstrated comparable kinematic data to traditional setups in laboratory environments their validity and reliability in constrained clinical spaces remain underexplored. Space limitations, such as narrow hallways common in clinical settings, could impact the accuracy of markerless motion capture data. The purpose of this study was to evaluate the validity and reliability of a markerless motion capture system implemented within a clinical hallway compared to a laboratory-based setup when estimating joint angles during clinical movement assessments among healthy adults. We analysed data from participants performing quiet standing, walking, and sit-to-stand activities in both environments on the same day. We hypothesized that there would be minimal differences between the data obtained from the constrained clinical hallway setup and the traditional laboratory-based setup, and that the reliability of measurements would remain high across both settings. Twenty-five healthy adults (34 ± 16 years, 15F) participated, reporting no clinical gait conditions and confirming a pain score ≤ three on a visual analog scale. Two camera setups were employed: an eight-camera laboratory configuration and a 10-camera constrained clinical hallway arrangement, both using Sony RX0 II cameras. The laboratory setup featured a circular layout with cameras spaced 2.7m apart along a 5.4m runway. The clinical setup used a 7.6m curved hallway, with cameras mounted on walls/ceilings to maximize visibility despite limited space. Each participant completed three tasks—quiet standing (30s), 60s self-paced walking, and five repetitions of sit-to-stand—twice at each site. The order of sites was counterbalanced, and tasks were performed on the same day, without guidance on clothing, to simulate clinical conditions. Video data were processed in Theia3D, using a generalized cross-validatory spline filter (8 Hz cutoff) and a six-degree-of-freedom knee model. Kinematic data was analysed in Visual3D, with comparisons made using Bland-Altman plots, Pearson correlations, and intraclass correlation coefficients (ICC(2, k)), evaluating both within-site and between-site reliability. The constrained clinical setup demonstrated high validity for joint angle estimates across tasks, with Pearson correlations >0.85 and RMSE < 2.33° for gait waveforms, except for hip flexion (2.33°). ICC values were excellent (≥0.85) for both within-site and between-site comparisons. Minimal differences were observed in discrete stance measures, with mean errors < 1.06° and limits of agreement < 5 °. Sit-to-stand results indicated higher variability, with wide limits of agreement (12–20°), consistent with previous findings in functional tasks. Markerless motion capture in constrained clinical spaces provides joint kinematics comparable to laboratory settings, supporting its potential integration into clinical workflows. This approach could enhance gait assessment accessibility by utilizing existing clinical spaces.
Ruder et al. (Tue,) studied this question.