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: 2023.1.0.3161 (patch 4).
Study Overview
Design
Fifteen children with cerebral palsy and 24 typically developing children underwent overground walking analysis with pelvis and lower limb kinematics recorded simultaneously by a marker-based and a markerless (Theia3D) system. GDI and GVI scores were calculated separately from each system’s kinematic data, using each system’s own normative reference data from the typically developing group, and compared between systems and between more and less affected limbs using two-way repeated-measures ANOVA.
Participants
Fifteen children with cerebral palsy (GMFCS levels I-IV; mean age 13.4 ± 3.7 years; 7 male, 8 female; a mix of hemiplegia and diplegia, with some using ankle-foot orthoses or wheeled walkers) and 24 typically developing children (mean age 10.5 ± 4.1 years; 13 male, 11 female) with no history of lower limb injury or conditions affecting gait.
Data Collection
Data were collected across two laboratory spaces using consistent settings, with marker-based and markerless (Theia3D) cameras recording simultaneously at 70 Hz. Participants wore 54 reflective markers for the marker-based system and completed five barefoot overground walking trials at a self-selected speed across a 5-meter walkway. Three children using wheeled walkers completed trials with their assistive device.
Analysis
Markerless video was processed through Theia3D and marker data through standard marker-based tracking, with two independently built skeletal models analyzed separately in Visual3D. GDI was calculated from 15 gait features (pelvic and hip angles in all three planes, knee flexion-extension, ankle dorsiflexion-plantarflexion, and foot progression angle) using each system’s own normative data from the typically developing group. GVI was calculated from nine spatio-temporal gait parameters. Differences in GDI and GVI between systems and limb impairment levels were tested using two-way repeated-measures ANOVA, with root mean square differences (RMSD) reported for individual kinematic waveforms.
Key Findings
- GDI scores were significantly and clinically meaningfully lower when measured with Theia3D. Theia3D-derived GDI averaged 6.9 points lower than marker-based GDI (65.63 vs. 72.49), a gap that exceeds the commonly cited 5-point threshold for a clinically significant GDI difference, meaning the same child’s gait could be scored as more pathological depending on which system generated the number.
- The system-based GDI gap was consistent regardless of how affected a child’s limb was. There was no interaction between system and limb impairment, meaning the roughly 7-point Theia3D-vs-marker-based offset applied similarly whether a limb was more or less affected, pointing to a systematic measurement difference rather than something specific to severity.
- GVI scores showed no significant difference between systems or limb impairment levels. Unlike GDI, this stride-to-stride variability measure was statistically comparable whether calculated from Theia3D or marker-based data, suggesting Theia3D can adequately support this particular type of clinical variability assessment.
- Four specific kinematic measures drove most of the disagreement between systems: pelvic tilt, hip flexion-extension, hip rotation, and foot progression angle. These showed RMSD values exceeding 10° between systems in both limbs of children with CP, while ankle kinematics showed the smallest differences of any sagittal- or transverse-plane measure.
- The authors explicitly recommend reporting the specific Theia3D version used in any published gait research involving the system. Based on their own experience across study populations, the authors note that kinematic results can shift meaningfully between Theia3D software versions, a practical point relevant to interpreting and comparing findings across the broader body of Theia3D validation literature.
What This Means for Research Teams
The authors frame this as an important caution for anyone using GDI across both systems: because switching from marker-based to Theia3D-derived GDI (or vice versa) could shift a score by more than the threshold considered clinically meaningful, they recommend the two systems’ GDI scores not be treated as interchangeable, particularly when tracking an individual child’s gait over time. GVI appears more robust to this difference, though the authors note their GVI calculation used only four strides per limb, just at the minimum recommended for assessing cycle-to-cycle variability, so they urge some caution there as well
The Role of Theia3D in Movement Analysis
Theia3D enabled the research team to capture full lower limb and pelvis kinematics in children with cerebral palsy, including several who used wheeled walkers or ankle-foot orthoses, without the marker placement burden that can be especially challenging in this population. The study also credits Theia3D’s high-resolution video and adequate capture volume coverage for making its spatio-temporal variability estimates (used in GVI) comparable to marker-based results, even where the pose-estimation-derived joint angles feeding into GDI diverged more substantially.
Study Limitations and Future Directions
The authors note their sample size was small in both groups, which may limit the ability to detect subtler differences in GDI and GVI beyond what was found. They used their own normative dataset, collected from the same typically developing children in their own lab using both systems, to calculate GDI and GVI for each system separately, a deliberate choice intended to control for differences in segment definitions between labs, though it does mean the results reflect this specific normative approach rather than comparison against another lab’s published normative values. The GVI calculation relied on only four strides, the minimum generally recommended for evaluating cycle-to-cycle gait variability in clinical applications, so the authors urge caution when interpreting GVI results, particularly with Theia3D. They also note their CP sample was predominantly diplegic (which typically shows less gait variability than hemiplegia), which may partly explain why no significant GVI differences were detected, and recommend future studies include larger samples that account for GMFCS level and assistive device use when examining kinematic differences between systems.
Citation: Poomulna, J., Knarr, B.A., Dutt, V., & Kingston, D.C. (2025). Comparison of gait deviation index (GDI) and gait variability index (GVI) measured by marker-based and markerless motion capture systems in children with cerebral palsy (CP). Gait & Posture, 115, 7-13.
Link: https://doi.org/10.1016/j.gaitpost.2024.10.018
Comparing Markerless and Marker-Based Capture for Pediatric Gait Research?
Standardized indices like GDI are only as useful as the consistency behind the number, and this study shows exactly where that consistency holds and where it needs a closer look between systems. If your research involves comparing GDI or GVI across capture systems in pediatric populations, talk to our team about the validation data available for your application.