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 v2023.1.0.3161, as stated in the paper.
Study Overview
Design
A within-subject study comparing lower-limb three-dimensional gait kinematics derived simultaneously from a marker-based system using the Plug-in Gait model and a markerless (Theia3D) motion capture system, in typically developing children, with a focus on how differences between the two systems varied across the gait cycle and by participant age.
Participants
Ten healthy children (8 boys, 2 girls; mean age 8.3 years; mean height 1.33 m; mean mass 31.4 kg), including three 6-year-olds, four 8-year-olds, two 10-year-olds, and one 13-year-old. Participants were able to walk independently without assistive devices, were free from musculoskeletal or neurological conditions affecting their gait, and had no prior history of lower-limb surgery.
Data Collection
A single, synchronized mixed-camera system combining 10 infrared marker-based cameras and 7 color video cameras captured all trials at 85 Hz within one shared calibration and reference frame. Retroreflective markers were placed on each participant's lower limb according to the Plug-in Gait model by a single, repeatability-tested assessor. After a static calibration trial, participants walked barefoot at a self-selected pace along a 10-meter walkway until five kinematic trials were collected. The same video data were processed simultaneously through Theia3D (default inverse kinematics model with 6 degrees of freedom at the pelvis, 3 at the hip, and 2 at the knee and ankle) and through the Plug-in Gait marker-based pipeline, with both outputs exported to Visual3D to generate two separate skeletal models for comparison.
Analysis
One left-sided gait cycle from each of the five trials was analyzed per participant, with gait events identified from foot marker kinematics and applied identically to both systems to ensure matched gait cycles. Generalised Additive Mixed Models were fitted to identify the systematic and random effects of measurement system, time across the gait cycle, and age, with approximate F-tests used to test for differences between systems while controlling for age and repeated measurements, using a Bonferroni-corrected significance threshold. Root-mean-square deviation (RMSD) was calculated for each participant's gait cycle to summarize pairwise differences between the marker-based and markerless systems, interpreted against a commonly used 5-to-10-degree guideline threshold for inter-assessor repeatability error in gait assessment laboratories.
Key Findings
- Significant differences and interactions between the Theia3D and Plug-in Gait models were found for most joint angles and planes of motion across the gait cycle and between age groups, with the statistical model explaining between 73% and 97% of the observed variation. Despite these significant differences, RMSD values were under 5 degrees for pelvic rotation and ankle dorsiflexion-plantarflexion specifically, indicating close agreement for these two measures.
- Pelvic tilt showed the largest difference of any measure (RMSD = 10.5°), with Theia3D consistently showing less anterior pelvic tilt across the entire gait cycle than the Plug-in Gait model, a pattern the authors note is comparable in magnitude to previously reported differences between other marker-based models and Theia3D in typically developing children.
- Hip rotation also showed a comparatively large difference (RMSD = 8.0°), with Theia3D showing more external hip rotation during the stance phase and less during initial swing compared with the Plug-in Gait model, a difference the authors note may reflect marker placement error in the marker-based data, systematic differences in the markerless system, or some combination of both, since accurate assessment of hip rotation is a well-known challenge for marker-based systems as well.
- The size of the difference between systems varied notably by age for several measures, including pelvic obliquity, hip and knee flexion-extension, hip abduction-adduction, and hip rotation, though there was no single age group that consistently showed larger or smaller differences across every joint. Ten-year-olds showed the smallest differences in pelvic obliquity and hip abduction-adduction, while the largest differences in hip flexion-extension and hip rotation were observed among both 8- and 10-year-olds, and 13-year-olds showed the greatest discrepancy specifically in knee flexion-extension.
- Systematic, largely consistent offsets between the two systems, rather than inconsistent tracking, appeared to explain much of the difference observed at the pelvis specifically, with the Plug-in Gait model defining a moderately anteriorly tilted pelvis compared with a more neutral pelvis position generated by the Theia3D model, a pattern the authors note is a common feature when comparing different marker-based models to one another as well, suggesting at least part of the pelvic tilt difference reflects differing model definitions rather than a markerless-specific error.
What This Means for Movement Science Researchers and Pediatric Assessment Teams
For researchers and teams who rely specifically on the Plug-in Gait model as their established reference, this study offers a direct, previously unavailable comparison point: Theia3D showed close agreement with Plug-in Gait for ankle motion and pelvic rotation, but pelvic tilt and hip rotation showed differences large enough that the two systems should not yet be treated as interchangeable for these specific measures. The finding that a child's age influenced the size of these differences, without following one single, consistent pattern across every joint, is a useful caution for teams planning markerless assessment across a wide pediatric age range, and reinforces the authors' recommendation that larger, more age-diverse samples are needed before drawing firm conclusions about how markerless kinematics perform across pediatric development.
The Role of Theia3D in Movement Analysis
Theia3D was the markerless motion capture system evaluated in this study, compared directly against the Plug-in Gait model, the most widely used marker-based biomechanical model in pediatric gait assessment, using data collected simultaneously from the same children. This specific comparison fills a gap in the existing evidence base, since prior Theia3D validation work in children had largely compared it against other marker-based modeling approaches rather than against Plug-in Gait directly, giving researchers and teams who use Plug-in Gait as their own reference a more directly applicable point of comparison.
Study Limitations and Future Directions
The authors note that the small sample size of ten children limits the generalizability of these findings, and that because gait characteristics can vary substantially with age, development, and individual variability, caution is warranted in extending these results to broader populations; they identify larger and more age-diverse samples as a priority for future work aiming to characterize these differences across developmental stages more precisely. The knee varus-valgus curves were assessed for crosstalk and fell under an established 10-degree threshold, though the presence of a knee valgus pattern during swing suggests some possibility of kinematic crosstalk remains. The authors also note that markerless kinematics generated by Theia3D may be influenced by factors related to the composition of its training dataset, including the age, sex, and ethnicity of the individuals represented in it, and that it is unlikely the underlying neural network was trained specifically on children wearing markers on their legs; because Theia3D is a proprietary system, researchers have limited visibility into how its training data may shape results. The authors identify larger studies involving movement-impaired pediatric populations as an important next step, noting that further investigation is needed to determine whether markerless lower-limb kinematics represent an accurate method in typically developing children before extending that conclusion to children whose gait differs meaningfully from a typical pattern.
Citation: Greaves, H., Eleuteri, A., Barton, G., Robinson, M. A., Gibbon, K. C., & Foster, R. J. (2025). Comparison of marker- and markerless-derived lower body three-dimensional gait kinematics in typically developing children. Sensors, 25(14), 4249. https://doi.org/10.3390/s25144249
Link: Read the full study
How Does Theia3D Compare Against the Plug-in Gait Model in Children?
This study directly compares Theia3D against the Plug-in Gait model, the most widely used marker-based biomechanical model in pediatric gait assessment, finding close agreement at the ankle and pelvic rotation alongside systematic differences at the pelvis and hip that varied somewhat by the child's age. If your work involves pediatric gait assessment and you want to understand how Theia3D compares against your own established marker-based model, talk to our team about applying Theia3D to your own pediatric research.