Study at a glance
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 v2024.2, as stated in the paper.
Study Overview
Design
A prospective, single-center, concurrent cohort study comparing shoulder kinematics derived simultaneously from a marker-based and a markerless (Theia3D) motion capture system during four activities of daily living, in participants with glenohumeral osteoarthritis and healthy controls, with a secondary aim of identifying which of four marker-based shoulder model configurations most closely matched markerless measurements.
Participants
One hundred participants: 85 with unilateral glenohumeral osteoarthritis scheduled for a shoulder replacement procedure, and 15 healthy controls with no known shoulder condition or current shoulder pain, included to help characterize normal shoulder motion ranges for comparison. Participants were between 40 and 80 years old and were excluded if they had a primary neurological condition, diabetic neuropathy, any non-shoulder-related condition impairing their ability to stand or walk without assistance, or were pregnant.
Data Collection
Participants with osteoarthritis were tested one week before their scheduled procedure; healthy controls were tested on a convenient date. Each participant was fitted with a full-body external reflective marker set along with several upper-extremity marker clusters for marker-based assessment (8 infrared cameras, University of Southampton Upper Limb Kinematic Model), while a markerless system (8 video cameras) processed through Theia3D captured the same trials simultaneously; both systems recorded at 100 Hz and were temporally synchronized. Participants completed four tasks bilaterally, each repeated three times: overhead reaching (lifting a weighted cup to a self-selected shelf height), drinking from a cup, hair brushing (three strokes), and a personal hygiene task (reaching behind the back toward the sacrum).
Analysis
Raw three-dimensional coordinates from both systems were smoothed using a fourth-order, zero-phase Butterworth low-pass filter. Because the activities involved combined shoulder flexion and abduction, the markerless system's output was compared against four variations of the marker-based upper-limb model, two flexion-based and two abduction-based configurations, to identify which configuration most closely matched markerless measurements. Peak angle and range of motion in flexion-extension, abduction-adduction, and internal rotation were compared between systems using intraclass correlation coefficients (ICC), Pearson correlation, and Bland-Altman analysis, with ICC values interpreted as poor (below 0.50), moderate (0.50-0.75), good (0.75-0.90), or excellent (above 0.90).
Key Findings
- Agreement between the markerless and marker-based systems was excellent for both peak angle and range of motion in flexion-extension and abduction-adduction across nearly every task, laterality, and model configuration tested (ICC > 0.90 in the large majority of comparisons), with markerless flexion-extension measurements generally falling within 10 degrees of the marker-based reference across all four activities.
- Internal and external shoulder rotation showed consistently weaker agreement across every task and model configuration, with most rotation comparisons falling into the poor-to-moderate range (ICC well below 0.75), and the markerless system tending to underestimate rotation range of motion throughout. The hair brushing and drinking tasks showed the strongest internal rotation agreement of the four tasks, though still considerably weaker than the flexion-extension and abduction-adduction results.
- Bland-Altman analysis showed little to no bias between the markerless system and marker-based measurements for flexion-extension and abduction-adduction, but showed a clear, systematic bias for internal rotation, consistent with the markerless system's tendency to underestimate rotation across tasks.
- As forward flexion and abduction angles approached the upper end of the shoulder's functional range of motion, agreement between the two systems weakened somewhat, though a consistent relationship between the systems remained present. This suggests markerless shoulder tracking may be somewhat less precise specifically near the extremes of shoulder range of motion.
- The specific marker-based model configuration used for comparison mattered: the two flexion-based model variants showed the closest overall agreement with the markerless system across most tasks, while the abduction-based variants showed comparatively weaker agreement for flexion-extension and abduction-adduction specifically during the overhead reaching task. This indicates that the choice of marker-based comparison model can meaningfully affect how markerless shoulder measurements appear to agree with a given reference.
What This Means for Movement Science Researchers and Upper-Extremity Assessment Teams
For researchers and teams working on upper-extremity motion analysis, this study offers a clear, specific picture: Theia3D showed excellent agreement with marker-based motion capture for shoulder flexion-extension and abduction-adduction across a range of everyday tasks and in a population with meaningful shoulder pathology, supporting its practical use for these specific planes of motion even outside a traditional gait laboratory. The consistent underestimation of internal and external rotation across every task and model configuration tested is an important, specific limitation to keep in mind: applications depending on accurate rotation measurement, evaluating rotator cuff function or assessing rotational range of motion loss, for example, will need to account for this known limitation until markerless rotation tracking is further refined. More broadly, this study's demonstration that markerless shoulder analysis performs comparably in a population with substantial shoulder pathology, not just in healthy individuals, is a meaningful step toward expanding motion analysis access beyond dedicated, fully equipped movement laboratories.
The Role of Theia3D in Movement Analysis
Theia3D was the markerless motion capture system evaluated throughout this study, compared directly against a marker-based reference system using four everyday upper-extremity tasks in both people with glenohumeral osteoarthritis and healthy controls. This represents one of the first direct evaluations of markerless shoulder kinematics specifically in a population with shoulder pathology, extending the evidence base for Theia3D beyond the lower-body gait applications most markerless validation research has focused on and into functional upper-extremity assessment, a domain where motion analysis has historically been underutilized relative to gait analysis.
Study Limitations and Future Directions
The authors note that while the markerless system was highly accurate across most of the shoulder's functional range of motion, agreement weakened somewhat as flexion and abduction angles approached their extremes, suggesting the system may be less sensitive, or less consistently validated, near the outer limits of shoulder range of motion. The authors also note that although no specialized clothing was required for the markerless system, differences in clothing between participants could still have affected measurement accuracy, since clothing was not standardized in this study. They identify future research investigating whether markerless shoulder accuracy varies based on individual participant characteristics, such as age, BMI, and osteoarthritis severity, as an important next step, and note more broadly that establishing dedicated, task-specific thresholds for what constitutes a practically meaningful difference in shoulder motion, since currently used thresholds vary considerably depending on the specific task and movement being assessed, would help translate these findings into clearer practical guidance.
Citation: Haddas, R., Morriss, N., Schillinger, E., Minto, J., Castle, P., Greif, D. N., Ramirez, G., Barber, P., Nicandri, G., Mannava, S., & Voloshin, I. (2025). Comparison of markerless and conventional marker-based shoulder kinematics models during activities of daily living in patients with glenohumeral osteoarthritis. Journal of Orthopaedic Research, e70047. https://doi.org/10.1002/jor.70047
Link: Read the full study
Does Markerless Capture Hold Up for Shoulder Motion During Everyday Tasks?
This study of 100 participants with and without glenohumeral osteoarthritis finds that Theia3D shows excellent agreement with marker-based motion capture for shoulder flexion-extension and abduction-adduction during everyday tasks, while consistently underestimating shoulder rotation. If your work involves upper-extremity motion analysis and you want to understand where markerless capture currently stands for shoulder assessment, talk to our team about applying Theia3D to your own upper-extremity research.