500 Section - Concurrent Validation of Markerless Motion Capture Systems Against VICON for Lower-Limb Biomechanics During Gait and Functional Tasks

Tuesday, May 19, 2026

6:00 PM - 7:30 PM PT

Presenting Author/Autor expositor(s)

YF

Yume Fukushima, BS

Graduate Student, MSc in Physical Therapy
Juntendo University, Graduate school Health Science, Department of Physical Therapy
Urayasu-shi, Chiba, Japan

Corresponding Author/Autor a quien dirigir la corr(s)

TM

Tadamitsu Matsuda, PhD

professor
Department of Physical Therapy, Faculty of Health and Medical Sciences, Juntendo University
Bunkyo-ku, Tokyo, Japan

Co-Author/Coautor(s)

KK

Kei Kakegawa, MS

Researcher
Juntendo University
Bunkyo-ku, Tokyo, Japan
JK

Junji Katsuhira, PhD

professor
Faculty of Social Welfare and Social Design, Department of Human Environment Design, Toyo University
Kita-ku, Tokyo, Japan
YT

Yuta Terada, BS

Graduate Student, MSc in Physical Therapy
Juntendo University, Graduate school Health Science, Department of Physical Therapy
Bunkyo-ku, Tokyo, Japan

Objectives : Technological advances have increased the clinical applicability of motion analysis, and markerless systems have emerged as accessible alternatives to laboratory-based marker-based capture. OpenCap estimates kinematics and kinetics from smartphone videos, while Theia3D provides rapid and accurate 3D pose estimation using synchronized video and deep learning. However, the accuracy of lower-limb kinematics and musculoskeletal model–derived joint moments from markerless systems relative to gold-standard measurements remains unclear. To our knowledge, this is the first study to simultaneously evaluate gait, stair ascent, and sit-to-stand tasks using two independent markerless systems (OpenCap and Theia3D) fully synchronized with the gold-standard marker-based system VICON and force plates. This study aimed to assess the agreement of lower-limb joint angles and model-based biomechanics outcomes to determine the clinical feasibility of markerless motion analysis.

Design: Twenty healthy adults performed level walking, sit-to-stand, and stair-ascent tasks. VICON, OpenCap, and Theia3D captured movements simultaneously at 60 Hz, with ground reaction forces collected using AMTI force plates. Joint angles in three planes and musculoskeletal model–based joint moments were computed and time-normalized. Waveform differences were evaluated using Statistical Parametric Mapping (SPM1D) paired t-tests.

Results: Sagittal plane kinematics showed strong agreement between systems, with no significant differences across hip, knee, or ankle during gait. Moderate discrepancies emerged in frontal and transverse planes, particularly in rotational and medial–lateral motions. Joint moments were similar within calibrated capture regions but demonstrated increased variability when movements extended beyond this space. OpenCap-estimated ground reaction forces resembled VICON under optimal camera positioning but tended to underestimate outside calibrated regions.

Conclusion: Markerless systems showed clinically acceptable agreement with VICON for sagittal plane biomechanics, especially in tasks with limited spatial progression. Accuracy decreased when gait extended beyond optimized capture regions. These findings support the practical utility and current limitations of markerless motion analysis in rehabilitation.