Study at a glance

In 10 children with achondroplasia and 10 children of average height, Theia3D showed good to excellent agreement with marker-based motion capture for most spatiotemporal gait parameters, but showed meaningful differences in several lower-limb kinematic measures for both groups, most notably an underestimation of pelvic tilt that the authors describe as a known limitation of earlier Theia3D versions.

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 v2023.3, as stated in the paper. This paper is an SSRN preprint and has not undergone peer review.

Study Overview

Design

A comparative cross-sectional study directly comparing spatiotemporal and three-dimensional lower-limb kinematic gait parameters derived simultaneously from a marker-based and a markerless (Theia3D) motion capture system, in children with achondroplasia and children of average height.

Participants

Ten children with achondroplasia (mean age 11.2 years; mean height 116.1 cm; mean weight 29.5 kg) and ten children of average height (mean age 10.4 years; mean height 147.4 cm; mean weight 36.9 kg), ages 8 to 13. All participants were able to walk independently without assistive devices for at least 10 meters. Participants were excluded if they had experienced a lower-limb trauma within the previous six months, and average-height participants were additionally excluded for any knee or ankle joint alignment concern. As expected given the defining features of achondroplasia, the two groups differed significantly in height, weight, and leg length, while sitting height did not differ significantly between groups.

Data Collection

Each participant walked multiple trials along a roughly 10-meter walkway at a self-selected pace, wearing minimal, skin-tight clothing and barefoot. Thirteen infrared Vicon Vue cameras recorded a 28-marker modified Plug-in Gait marker set at 200 Hz, while eight synchronized FLIR video cameras simultaneously recorded the same trials at 100 Hz for markerless processing through Theia3D (v2023.3, default settings, three-degree-of-freedom knee model). Five valid trials containing a full gait cycle for both legs were retained per participant for analysis.

Analysis

Marker-based data were processed in Vicon Nexus and Polygon with gait events set manually by trained staff; markerless data were processed through Theia3D and a customized analysis script, with spatiotemporal parameters calculated using an established foot-kinematics-based event-detection method. Both datasets were filtered with a second-order low-pass Butterworth filter (5 Hz cutoff) and time-normalized to 101 points per gait cycle. Between-group anthropometric differences were assessed using the Wilcoxon rank test and Pearson's r. Overall between-system differences in kinematic waveforms were assessed using statistical nonparametric mapping with a t-test, spatiotemporal agreement was assessed using Bland-Altman plots and Spearman rank correlation, and kinematic inter-system variability was assessed using median absolute error, interpreted against an established threshold in the gait literature under which errors of 2 to 5 degrees are considered relevant to interpretation and errors above 5 degrees are considered large enough to be potentially practically significant.

Key Findings

  • Most spatiotemporal gait parameters showed good to excellent agreement between the marker-based and markerless systems in both groups. Spearman correlations ranged from 0.72 to 0.98 for children with achondroplasia and 0.61 to 0.98 for children of average height across parameters including gait velocity, cycle time, and stride and step length.
  • Step width was the one spatiotemporal parameter that showed consistently poor agreement between systems in both groups (Spearman rs = 0.11 for achondroplasia, 0.10 for average-height children), alongside modestly larger Bland-Altman differences for stance time, swing time, and double limb support time in both groups.
  • Pelvic tilt showed statistically and practically meaningful differences between systems across the entire gait cycle in both groups, with Theia3D consistently estimating greater posterior pelvic tilt than the marker-based reference (median absolute error 11.56 degrees for achondroplasia, 7.65 degrees for average-height children). The authors describe this pattern as a previously identified limitation of earlier Theia3D software versions.
  • Several other kinematic measures also showed practically meaningful differences between systems, with the specific pattern differing somewhat by group. For children with achondroplasia, hip flexion-extension, transverse-plane knee rotation, and transverse-plane pelvic rotation each showed median absolute errors in the practically meaningful range (6.13 to 9.07 degrees), alongside sagittal- and coronal-plane knee differences and all three ankle planes. For children of average height, transverse-plane pelvic, knee, and ankle rotation also showed practically meaningful differences, alongside the pelvic sagittal-plane finding above.
  • Theia3D showed notably greater variability specifically in its transverse-plane pelvic rotation estimates, which the authors interpret as reflecting difficulty in the software's ability to confidently resolve pelvic position and rotation, and noted a kinematic timing shift specific to the achondroplasia group's gait curves, plausibly related to how the event-detection approach handled this group's more variable gait pattern, since event detection remained consistent for the average-height group.

What This Means for Movement Science Researchers and Care Teams

For researchers and teams planning gait assessment in children with achondroplasia or other populations with atypical body proportions, this study offers an early, directly relevant data point rather than a final answer: Theia3D reliably captured the overall timing structure of the gait cycle and showed strong agreement for most spatiotemporal parameters in both groups tested, supporting its practical value for measures like walking speed and step length, but several joint angles, pelvic tilt in particular, showed differences large enough that absolute kinematic values should be interpreted with caution in the software version evaluated here. The authors note that newer Theia3D versions are reported to address the pelvic tilt estimation pattern identified in this study, making a version-to-version comparison a useful next step for anyone extending this work, and underscoring that reliability findings in rapidly developing markerless software can be specific to the version tested.

The Role of Theia3D in Movement Analysis

Theia3D was the markerless motion capture system evaluated in this study, assessed directly against a 28-marker Vicon-based reference system in two groups: children with achondroplasia and children of average height. This head-to-head design allowed the authors to establish not only how well Theia3D agreed with marker-based motion capture overall, but whether that agreement held consistently across a population with substantially different body proportions than those typically represented in markerless validation research, a direct and practically relevant test of the software's generalizability.

Study Limitations and Future Directions

The authors note that achondroplasia's rarity limited the achievable sample to ten participants per group, and that recruitment was further limited by the distance some families needed to travel and the fact that not every family visiting the children's hospital also visited the movement laboratory on the same trip. The resulting small sample size, combined with a non-normal data distribution, required nonparametric statistical methods, which the authors note complicates direct comparison with prior literature that has typically relied on parametric reliability statistics such as the intraclass correlation coefficient. Because data collection and processing took place over an extended period, this study used Theia3D version 2023.3 rather than a more current release; the authors note that newer software versions are reported to have addressed the pelvic tilt estimation pattern identified here, and recommend that future work compare outcomes across software versions directly. Gait event detection relied on a customized script calibrated to this study's specific, relatively small dataset, and the authors suggest that a larger sample could support more robust event-detection performance. The authors also note that a widely used alternative analysis platform was not a feasible option for this study's workflow, leading them to adopt a customized scripted approach instead.

Citation: Hergenröther, M., Palm, K., Mohnike, K., & Witte, K. (2026). A comparison of the lower body gait parameters between a markerless and marker-based motion capture system in children with achondroplasia. SSRN Preprint. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6233089

Link: Read the full preprint

How Does Markerless Capture Perform in Children With Achondroplasia?

This preprint offers an early, direct comparison of Theia3D against marker-based motion capture in children with achondroplasia and children of average height, finding strong agreement for most spatiotemporal gait parameters alongside specific kinematic differences, most notably in pelvic tilt, that the authors attribute in part to a known limitation of the software version tested. If your work involves gait assessment in populations with atypical body proportions and you want to understand where markerless capture currently stands, talk to our team about the latest Theia3D performance for your specific research population.

Recent Posts