Human Movement & Mobility Research
Qualisys Miqus

A Comparison of Lower Body Gait Kinematics and Kinetics Between Theia3D Markerless and Marker-Based Models in Healthy Subjects and Clinical Patients

Scientific Reports
•
2024
•
Original research
Theia3D version:
v2022.1.0.2309
More Articles

Study at a glance

This study compared Theia3D markerless gait analysis to a marker-based system in 12 healthy individuals and 34 clinical patients with orthopaedic and neuro-orthopaedic conditions, finding generally similar gait patterns overall but non-comparable results for hip and knee rotation, including a clinical case where the two systems led to different diagnostic interpretations for the same patient.

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

Twelve healthy individuals and 34 clinical patients with orthopaedic or neuro-orthopaedic conditions underwent gait analysis with Theia3D markerless and a marker-based (CAST model) system recording simultaneously, in an active clinical gait laboratory setting. Kinematics and kinetics were compared between systems statistically, and clinical interpretation was separately assessed by comparing individual patients’ data against system-specific normative reference bands, including a detailed case study.

Participants

Twelve healthy individuals (4 male, 8 female, ages 8 to 61) free of neuro-orthopaedic disease, and 34 clinical patients (24 male, 10 female, ages 7 to 55) referred for clinical gait analysis, with diagnoses comprising 64% neuro-orthopaedic disorders (including cerebral palsy and hereditary motor sensory neuropathy) and 36% orthopaedic conditions (including flat feet). Twenty-nine of the 34 patients were bilaterally affected.

Data Collection

Marker-based data were captured using 11 wall-mounted infrared cameras and the CAST marker model, with three force plates recording ground reaction forces. Markerless data were captured simultaneously using 8 Theia3D video cameras, borrowed from the marker-based system’s manufacturer for a four-week loan period and synchronized with the existing marker-based setup. Participants walked barefoot at a self-selected pace over a 13-meter walkway, completing six walking trials each.

Analysis

Marker-based data were processed in Visual3D using a six-degree-of-freedom segment model, while markerless video was processed through Theia3D’s own inverse kinematics before export to Visual3D for comparison. Joint angles and moments were compared between systems using RMSE, Pearson’s correlation coefficients, and one-dimensional statistical parametric mapping, with a joint angle considered non-comparable between systems when RMSE exceeded 5° alongside very low correlation. Inter-trial variability was also calculated for both systems to account for natural stride-to-stride variation.

Key Findings

  • Most kinematic and kinetic measures showed good agreement between systems, in both healthy and clinical populations. Knee and ankle flexion, hip flexion and adduction, and foot progression angle all showed relatively low RMSE and high correlation, and joint moments and power showed good agreement overall (RMSE 0.1-0.4 Nm/kg, correlation generally above 0.8).
  • Hip and knee internal/external rotation were clearly non-comparable between systems. Both measures showed RMSE around 9° combined with correlation coefficients near zero or negative in both healthy and clinical groups, meaning the two systems sometimes disagreed not just on magnitude but on direction of rotation.
  • Pelvic anterior tilt was significantly underestimated by the markerless system, with downstream effects on hip flexion. This systematic offset propagated into the hip joint, reducing measured hip flexion, and the authors note that because Theia3D’s underlying model isn’t open-source, it’s difficult to pinpoint exactly why without further study specifically focused on pelvis definition differences.
  • A detailed clinical case study found the two systems produced genuinely different clinical interpretations for the same patient. In a patient with hereditary motor sensory neuropathy and a severe drop foot, the marker-based system captured a large, clinically meaningful compensatory external hip rotation used to clear the ground during swing, a finding confirmed by video review, while the markerless system instead showed a slight internal rotation that did not reflect this compensation at all.
  • Across nearly all patients studied, hip and/or knee rotation produced different clinical outcomes between systems. The authors report that of all patients examined for this comparison, all but three showed differing interpretations in hip and/or knee rotation between the markerless and marker-based results, underscoring that this is not an isolated case but a consistent pattern across the clinical sample.

What This Means for Gait Research Teams

This study’s findings point to a nuanced picture for gait research applications: sagittal-plane measures such as knee and ankle flexion, hip flexion, and foot progression showed strong agreement with marker-based data, while hip and knee transverse-plane rotation showed much larger discrepancies, including directionally opposite results in some cases. The authors caution that relying on markerless rotation data alone could produce a different read on a patient’s compensatory gait strategy than marker-based analysis would, as the case study above illustrates, and recommend extensive further validation before any clinical application.

The Role of Theia3D in Movement Analysis

Theia3D allowed the research team to capture gait kinematics and kinetics in both healthy volunteers and real clinical patients without marker placement, reducing preparation time and avoiding the marker loss or occlusion issues that marker-based systems can experience, particularly relevant for pediatric patients and those with mobility challenges in a busy clinical gait lab. At the same time, the study is candid that markerless capture is not yet a drop-in replacement for marker-based systems across every measure, particularly transverse-plane hip and knee rotation, where the underlying differences in how each system defines its model appear to matter clinically.

Study Limitations and Future Directions

The authors emphasize that marker-based motion capture itself is not a validated gold-standard benchmark, true accuracy for either system could only be established against a reference like bi-planar videography, and marker-based systems carry their own known sources of error, including inter-operator variability in marker placement and soft tissue artifact, which likely contributed to some of the spread seen in the marker-based data too. They note that because Theia3D’s underlying model is proprietary and not open-source, it is difficult to pinpoint exactly why specific differences (like the pelvic tilt offset) occur, and recommend future studies focus specifically on investigating pelvis model definitions. The authors also point out that comparisons across different published Theia3D validation studies should be made cautiously, since studies have used different versions of the software, different marker-based comparison models, and different sample sizes, all of which can affect results independent of any change in the technology itself.

Citation: D’Souza, S., Siebert, T., & Fohanno, V. (2024). A comparison of lower body gait kinematics and kinetics between Theia3D markerless and marker-based models in healthy subjects and clinical patients. Scientific Reports, 14, 29154.

Link: https://doi.org/10.1038/s41598-024-80499-8

Know Exactly Where Markerless Capture Fits in Your Gait Lab

This study offers a detailed picture of where markerless gait analysis closely tracks marker-based data and where meaningful differences remain, particularly for hip and knee rotation. If your research involves comparing gait analysis systems, talk to our team about the validation data available for your specific application

Recent Posts