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 used: v2022.1.0.2309.
Study Overview
Design
Twenty-two healthy adults completed two gait analysis sessions roughly a week apart, each session including walking trials in both tight-fitting and loose, everyday clothing. Theia3D markerless kinematics were calculated for each trial, and inter-trial, inter-session, and inter-session-clothing variation were quantified, along with direct root mean square differences between the two clothing conditions.
Participants
Twenty-two healthy adults (8 female, 14 male; mean age 25 ± 4 years; height 174 ± 9 cm; weight 70 ± 11 kg), free from lower-limb injury and not undergoing rehabilitation or affected by neuromuscular or musculoskeletal impairment.
Data Collection
Each participant attended two sessions, with the second within 14 days of the first (average 6 days apart). In each session, participants completed ten overground walking trials at a self-selected pace, five in tight-fitting clothing and five in their own loose, everyday clothing, recorded using eight synchronized Theia3D video cameras at 180 Hz. Participants wore the same shoes across both sessions but were not required to bring the same clothing.
Analysis
Video data were processed through Theia3D’s default full-body model and exported to Visual3D for gait cycle extraction, using a single right-sided gait cycle per trial to represent fully independent walking trials rather than multiple strides within one trial. Inter-trial variation was calculated from the five same-day, same-clothing trials; inter-session variation from trials across both days in the same clothing type; and inter-session-clothing variation across both days and both clothing types combined. Root mean square difference was calculated separately to quantify the direct difference between clothing conditions.
Key Findings
- Inter-trial variation stayed under 2° on average, consistent with thresholds considered acceptable for gait analysis generally. Peak inter-trial variation reached 3.84° at the knee and 2.86° at the ankle in the sagittal plane, both still within commonly cited acceptable ranges.
- Inter-session variation (testing a week apart) stayed under 3° on average, adding only about 1° of additional variation beyond inter-trial variation alone. This indicated that returning for a second session increased total variation by roughly 50%, a meaningfully smaller increase than typically seen when a second human assessor is introduced in marker-based testing.
- Clothing choice made essentially no measurable difference to reliability. Loose clothing showed only marginally more inter-trial variation (1.35° vs. 1.29°) and inter-session variation (2.00° vs. 1.88°) than tight clothing, differences the authors characterize as negligible, and the root mean square difference between clothing conditions stayed under 2° for every joint and plane measured.
- Pelvis kinematics showed the smallest variation of any body segment studied, the first time this has been specifically evaluated for Theia3D. The authors suggest this may be because Theia3D models the pelvis as an independent six-degree-of-freedom segment, making it less dependent on the pose estimation of other body segments in the kinematic chain, though they note this explanation isn’t confirmed.
- Overall variation was lower than in several earlier Theia3D reliability studies, which the authors tentatively attribute to software version. This study used a newer Theia3D version than prior reliability research, and the authors explicitly recommend that all Theia3D studies report the specific software version used, since updates to the underlying algorithm appear capable of measurably changing reliability results over time.
What This Means for Gait Labs and Researchers
For gait labs weighing whether markerless capture is practical enough for routine use, this study offers a specific, actionable answer to a common logistical question: participants do not appear to need special tight-fitting clothing for reliable Theia3D gait analysis, their own everyday loose clothing performs comparably well. The authors also point out a structural advantage worth considering for labs running multi-assessor marker-based protocols: because markerless capture doesn’t depend on a human placing markers, it removes inter-assessor variation entirely, a source of variability the cited marker-based literature places well above the inter-trial and inter-session variation reported here. They are careful to note, however, that this reliability finding is specific to healthy adults and this particular software version, and call for the same evaluation to be extended to clinical populations and children before drawing conclusions there.
The Role of Theia3D in Movement Analysis
Theia3D allowed the research team to assess gait reliability across sessions and clothing conditions without requiring markers or the preparation time and clothing constraints that come with marker-based testing, and to evaluate pelvis kinematics specifically for the first time in markerless reliability literature. The authors highlight that because markerless output doesn’t depend on a human assessor’s marker placement, it structurally avoids the inter-assessor variation that is a recognized, substantial source of error in multi-assessor marker-based gait labs.
Study Limitations and Future Directions
The authors note some uncontrolled variation in clothing within the study itself, since several participants’ choice of “loose” everyday clothing happened to be shorts and t-shirts with exposed knees and elbows, and clothing contrast wasn’t standardized beyond asking participants to wear contrasting colors between top and bottom. Only right-side gait data were analyzed, and the results are specific to the Theia3D software version used in this study, with further updates having been released since the analysis was conducted. Walking speed was not controlled between sessions or clothing conditions, which could have contributed some additional variation to the results. The authors also note that more challenging clothing types, such as long robes, skirts, or dresses, remain untested, even though Theia3D has separately been shown to cope well with additional equipment like backpacks in other published research.
Citation: Augustine, S., Foster, R., Barton, G., Lake, M.J., Sharir, R., & Robinson, M.A. (2025). The inter-trial and inter-session reliability of Theia3D-derived markerless gait analysis in tight versus loose clothing. PeerJ, 13, e18613.
Link: https://doi.org/10.7717/peerj.18613
Could Your Gait Lab Skip the Tight-Clothing Requirement?
This study suggests participants’ own everyday clothing may be reliable enough for markerless gait analysis, without the preparation and comfort tradeoffs of requiring tight-fitting attire. If you’re evaluating how markerless capture could simplify your gait analysis protocol, talk to our team about what this reliability research means for your lab’s setup.