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 v2023.1.0.310, as stated in the paper. Here, Theia3D served as the markerless gait-analysis tool whose own internal measurement reliability was being characterized, rather than as a system being validated against marker-based motion capture within this particular study.
Study Overview
Design
A test-retest reliability study examining how the number of gait strides averaged together affects the reliability and measurement error of markerless (Theia3D) lower-limb kinematics and kinetics in older adults, including scenarios with an unequal number of strides averaged across two sessions.
Participants
Twenty-nine healthy older adults (13 women, 16 men; mean age 75.2 years).
Data Collection
Participants completed gait assessments on two occasions, roughly nine days apart, walking at a self-selected speed in their own clothing and footwear across a 12-meter laboratory walkway for approximately 20 passages per session. Movement was captured using 8 video cameras at 85 Hz, synchronized with 3 floor-embedded force plates, with video data processed through Theia3D and lower-limb joint angles and internal moments calculated in Visual3D.
Analysis
Twenty time-normalized strides were extracted from each session, and non-consecutive random subsets ranging from 2 to 19 strides were repeatedly selected and averaged within each participant and session, including scenarios with an unequal number of strides between the two sessions. Reliability was quantified using the integrated intraclass correlation coefficient and standard error of measurement for full joint angle and moment waveforms, and the intraclass correlation coefficient with 95% confidence intervals for discrete range-of-motion and peak values. A minimally acceptable stride number was defined as the smallest number at which reliability deflected less than 10% from its maximum value, combined with an absolute reliability threshold of 0.75.
Key Findings
- Reliability stabilized, deflecting less than 10% from its highest achievable value, once 7 or more strides were averaged with an equal number of strides contributing to both sessions, reaching reliability values between 0.77 and 0.92 across joint angles and moments.
- Transverse-plane movement, rotational motion at the hip and ankle in particular, required the highest number of strides to reach stable reliability, consistent with prior research finding this plane of motion more variable than sagittal-plane motion for both markerless and marker-based gait analysis.
- Measurement error for kinematics rarely exceeded a 2-degree threshold considered suitable for practical interpretation of gait, and generally 4 strides were sufficient to bring joint moment error within an acceptable 5% threshold, with a small number of specific measures, notably hip transverse-plane moments, requiring more strides to reach that error threshold.
- Using fewer strides in just one of the two sessions had less impact on reliability than reducing the number of strides equally in both sessions, meaning a shorter follow-up assessment, for example when someone has less time or tolerance on a second visit, does not necessarily compromise the ability to reliably detect change over time.
- For specific, commonly reported gait parameters, including hip, knee, and ankle range of motion and peak ankle plantarflexion angle and moment, reliability with narrow confidence intervals was generally achieved with as few as 2 to 3 strides, somewhat fewer than the 7-stride threshold needed for full waveform-level reliability across all measures.
What This Means for Movement Science Researchers and Older-Adult Assessment Teams
For researchers and teams designing gait assessment protocols for older adults or other populations with limited assessment tolerance, this study offers a direct, actionable finding: a minimum of about 7 strides is generally sufficient to achieve reliable markerless kinematics and kinetics, considerably fewer than the 10 strides previously recommended for other motion capture approaches. The additional finding that shortening a follow-up session does not meaningfully compromise reliability adds further practical flexibility, supporting protocols that adapt to how much walking a given individual can comfortably complete on a given day without sacrificing the ability to track change over time. Together, these findings support more efficient, less burdensome markerless gait assessment protocols specifically tailored to populations, like older adults, where assessment time and tolerance are meaningful practical constraints.
The Role of Theia3D in Movement Analysis
Theia3D served as the markerless motion capture system whose own internal test-retest reliability was the direct subject of this study; rather than comparing Theia3D-derived measurements against marker-based motion capture, the researchers examined how the amount of markerless data collected, specifically, the number of gait strides averaged, affected the consistency of its own output. This kind of methodological reliability research is a necessary complement to validation studies, since even a measurement tool shown to agree closely with a marker-based reference still requires enough data per assessment to produce a stable, trustworthy result.
Study Limitations and Future Directions
The authors note that their random stride-selection approach could be seen as a limitation, though a prior pilot analysis using sequential rather than randomly selected stride subsets produced comparable stability, suggesting their reliability estimates are not strongly dependent on how strides were selected. The study sample was limited to healthy older adults, so these specific stride-number thresholds may not directly generalize to other populations with different movement characteristics or gait variability, including those with movement-related differences that could themselves affect stride-to-stride consistency. The authors identify extending this kind of stride-number reliability analysis to other populations and gait conditions as a useful next step, and note that their findings, particularly around the flexibility of using an unequal number of strides between sessions, may be especially valuable for simplifying future longitudinal study designs in populations with limited tolerance for lengthy assessments.
Citation: Carvalho, A., Vanrenterghem, J., Pataky, T. C., Robinson, M. A., Veloso, A. P., & Moniz-Pereira, V. (2026). How many strides are needed for reliable markerless gait analysis? Gait & Posture, 123, 110020. https://doi.org/10.1016/j.gaitpost.2025.110020
Link: Read the full study
How Many Strides Does Reliable Markerless Gait Analysis Actually Require?
This study of 29 healthy older adults finds that averaging as few as 7 gait strides was generally sufficient for reliable Theia3D-derived kinematics and kinetics, a practical finding for streamlining markerless gait assessment in populations with limited tolerance for lengthy walking trials. If your work involves gait assessment protocol design, talk to our team about applying Theia3D to your own research.