Study at a glance

Across 486 participants tested at three independent orthopaedic assessment sites in Canada, Theia3D consistently detected the gait deviations associated with knee osteoarthritis at every site, with site-to-site measurement differences smaller than those previously reported for marker-based motion capture across multiple sites, despite no extensive cross-site standardization effort.

Summary

Disclaimer: The following summary details independent academic research. Theia3D is intended for research, sports performance and educational use only. It is not cleared as a medical device and is not intended to be used for the diagnosis, treatment, mitigation or prevention of any disease, injury or medical condition.

Editor's note: Theia3D version v2023.1.0.3161, as stated in the paper.

Study Overview

Design

A multicentre study comparing gait kinematics and spatiotemporal parameters between individuals with knee osteoarthritis and asymptomatic participants, collected independently at three distinct orthopaedic assessment sites across Canada using a shared markerless motion capture system, with a secondary aim of examining site-to-site variation in the underlying kinematic measurements.

Participants

Four hundred eighty-six participants across three sites: 351 with knee osteoarthritis and 135 asymptomatic participants recruited as a convenience sample at each site. Individuals with knee osteoarthritis were included based on a diagnosis of end-stage disease in at least one knee and were awaiting a consultation or scheduled for knee joint replacement; diagnoses were made by trained orthopaedic assessment staff at each site based on symptom presentation and radiographic findings. All participants provided written informed consent, including data-sharing consent approved by each site's research ethics board.

Data Collection

Testing took place within public hospitals at each site's established orthopaedic assessment service, using the same camera model (Sony DSC-RX0M2) at each location, though the number of cameras (8 to 10) and capture volume dimensions varied by site based on available space. Participants completed a self-selected-pace walking task, walking continuously back and forth along each site's walkway for approximately 60 seconds or 6 lengths, wearing their own clothing and footwear with loose or additional garments removed where possible. All video was processed using the same version of Theia3D (v2023.1.0.3161) with matched processing preferences across sites, generating hip, knee, and ankle joint angles and spatiotemporal gait parameters for each participant.

Analysis

A minimum of 8 strides per participant was required for inclusion, based on established stride-count thresholds for stable repeatability in healthy older adults. One knee was analyzed per participant, with the most severe knee chosen for osteoarthritis participants. Principal component analysis was used to characterize joint angle waveform differences in the sagittal plane of the hip, knee, and ankle, and the frontal plane of the knee, with the resulting components representing the majority of overall waveform variation. One-way and two-way linear models examined group and site differences in demographics, spatiotemporal metrics, and joint angle principal component scores. Site-to-site kinematic variability was further quantified by comparing each site's group-specific mean waveform to the pooled grand mean, using root-mean-square, absolute maximum, and standard deviation of the resulting differences, and contextualized against previously reported inter-site measurement error thresholds from marker-based research in healthy samples.

Key Findings

  • Individuals with knee osteoarthritis showed a clear, consistent set of gait deviations compared with asymptomatic participants, detected reliably at all three sites. The osteoarthritis group walked significantly slower (0.95 vs 1.35 m/s), with a slower cadence, longer stance time, and shorter stride length (all P<0.0001), along with reduced hip and ankle range of motion, an earlier and lower peak knee flexion angle, and a higher knee adduction angle during stance compared with asymptomatic participants.
  • Site-to-site kinematic differences were small relative to the grand mean, with root-mean-square values generally below 2 degrees across the hip, knee, and ankle, and standard deviations of these differences consistently below 1.3 degrees, indicating that observed site-level variability reflected small, largely consistent offsets rather than meaningfully different measurement patterns.
  • These site-to-site differences fell within or below previously reported inter-site measurement error thresholds from marker-based motion capture studies in healthy samples (approximately 3.5 degrees for hip flexion, 4 degrees for knee flexion, 2 degrees for knee adduction, and 2.5 degrees for ankle dorsiflexion), and were notably smaller than the roughly 2 to 4 times larger inter-site differences reported in one of the few prior marker-based multicentre comparisons specifically in a knee osteoarthritis population.
  • Centre-specific effects were detected in several joint angle components, generally reflecting small, consistent offsets in overall magnitude or timing rather than disagreement in the underlying group-level pattern, with the authors attributing these largely to differences in camera placement, capture space dimensions, and participant demographics across sites rather than to inconsistency in the markerless system itself.
  • Two of the three sites collected data outside of a traditional, dedicated gait laboratory space, using hallways within their orthopaedic assessment areas, and still detected the same consistent osteoarthritis-related gait deviations as the dedicated laboratory site, demonstrating that markerless motion capture's reduced setup and space requirements translated into practical, workable data collection in non-laboratory assessment settings.

What This Means for Movement Science Researchers and Multi-Site Research Teams

For researchers and teams planning multi-site biomechanical research, this study offers direct, large-sample evidence that a shared markerless motion capture system can produce consistent, pooled results across independently operated sites, without the extensive standardization protocols that marker-based multi-site studies typically require. The specific finding that markerless site-to-site variability was smaller than previously reported marker-based benchmarks, even when sites used different camera counts, capture volumes, and testing spaces including hospital hallways, is a particularly practical result for teams weighing the feasibility of expanding a single-site study into a multi-site one. More broadly, this study offers one of the largest and most heterogeneous knee osteoarthritis gait samples assembled using markerless motion capture to date, providing a meaningful foundation for future large-scale biomechanical research and for the broader goal of integrating movement analysis into varied care settings.

The Role of Theia3D in Movement Analysis

Theia3D was the markerless motion capture system used at all three sites in this study, enabling independently operated research sites to collect and pool gait data using a shared measurement approach without extensive cross-site standardization. This multicentre design directly tested whether markerless motion capture's practical advantages, reduced setup time, no marker placement, and flexibility in testing space, translate into genuinely comparable data when the same software is deployed by different teams in different real-world settings, building on prior work establishing Theia3D's individual-site reliability and extending that evidence specifically to the multi-site context.

Study Limitations and Future Directions

The authors note that these findings are specific to Theia3D and may not extend to markerless systems with different levels of validation or maturity, and that only one camera model was used across sites, though differences in camera specifications such as resolution and lens could still influence results with other hardware. Minor protocol differences between sites, including trial number, order, and duration, may have contributed to some inter-site variability, though the authors note the site that collected more strides per participant likely had minimal impact on results. Age differences were present among asymptomatic participants across sites, which may have contributed to some of the observed inter-site biomechanical variability. The authors also note that osteoarthritis inclusion was based on diagnosis rather than standardized, research-oriented criteria such as radiographic grading or structured outcome measures, which may have introduced some heterogeneity in disease severity across the sample; they note, however, that this heterogeneity reflects real-world populations and is important for building broadly applicable models. The authors identify future research investigating the specific impact of camera placement on biomechanical outcomes as a useful next step for further refining best practices as multi-site markerless efforts continue to expand.

Citation: Outerleys, J., Laende, E., Malek, M., Civiero, S., Madden, K., Ruder, M., Dunbar, M., Adili, A., Kobsar, D., Wilson, J., & Deluzio, K. (2025). Clinical integration of markerless motion capture: A multicentre study of gait in knee osteoarthritis. Journal of Biomechanics, 192, 112952. https://doi.org/10.1016/j.jbiomech.2025.112952

Link: Read the full study

Can Markerless Motion Capture Deliver Consistent Results Across Multiple Sites?

This study pools data from 486 participants across three independently operated Canadian orthopaedic assessment sites, finding that Theia3D consistently detected knee osteoarthritis-related gait deviations at every site, with smaller site-to-site measurement differences than previously reported for marker-based motion capture. If your work involves multi-site biomechanical research or integrating gait assessment into varied real-world settings, talk to our team about applying Theia3D to your own multicentre research.

Recent Posts