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: The paper states the Theia3D version used as v2023.1.0.310; this may be a truncated or abbreviated rendering of a fuller version number such as v2023.1.0.3161, but is reproduced here exactly as it appears in the source text.
Study Overview
Design
A within-subject study directly comparing joint kinematics and kinetics derived simultaneously from a marker-based and a markerless (Theia3D) motion capture system during overground walking in children with cerebral palsy.
Participants
Fifteen children diagnosed with cerebral palsy (11 boys, 4 girls; mean age 13.66 years; mean body mass 44.50 kg; mean height 1.49 m), presenting with a range of gait deviation patterns including crouch gait, true equinus, and apparent equinus. As most participants were affected bilaterally, only the right limb was considered for analysis.
Data Collection
Assessments used a standard marker-based system (10 infrared cameras, 36 retroreflective markers placed using the Calibrated Anatomical Systems Technique) and a markerless system (8 color video cameras processed through Theia3D), both synchronized and recording at 85 Hz, along with three floor-embedded force plates recording at 850 Hz. A static calibration trial in anatomical standing position was recorded by both systems before anatomical markers were removed to avoid interfering with natural movement. Participants then walked along a 10-meter path at a self-selected speed, stepping over the centrally located force plates, with at least one foot required to fully contact a force plate for a trial to be included; data collection continued until each participant completed 10 successful trials, with rest breaks provided as needed.
Analysis
Marker-based data were processed using a dedicated gait analysis software module, with gait events automatically defined from force plate data and a biomechanical model built from the static calibration trial. Markerless data were processed through Theia3D's default inverse kinematics pipeline (6 degrees of freedom at the pelvis, 3 at the hip, knee, and ankle), smoothed with a generalized cross-validatory spline filter, and exported for analysis; 10 gait cycles per participant were included. Kinematic and kinetic variables were averaged and compared between systems using root-mean-square difference (RMSD) and two-tailed paired-sample t tests, along with statistical parametric mapping to assess waveform differences across the entire gait cycle rather than only at discrete time points.
Key Findings
- Sagittal-plane joint kinematics were highly comparable between the two systems, with knee flexion angle showing particularly good alignment and agreement throughout both the static and walking trials (RMSD = 5.0°), a pattern the authors note may be partly explained by participants wearing shorts during testing, which kept the knee area visible to the markerless cameras without any markers placed there during walking.
- Hip flexion and pelvic tilt showed a statistically significant difference throughout the entire gait cycle, with root-mean-square differences exceeding 10 degrees for both measures (hip flexion RMSD = 10.68°; pelvic tilt RMSD = 11.16°), a pattern the authors attribute largely to a consistent, systematic offset between how the two systems define neutral pelvis position and hip flexion, rather than to inconsistent tracking, meaning the underlying waveform shapes were similar even though absolute values differed.
- Frontal-plane agreement was comparatively better than transverse-plane agreement but still showed meaningful differences, with significant differences observed specifically at hip adduction and pelvic obliquity during early swing, while transverse-plane agreement was the weakest of the three planes assessed, with root-mean-square differences exceeding 10 degrees at the hip, knee, and ankle (hip = 16.22°, knee = 12.19°, ankle = 10.86°), except at the pelvis, where transverse-plane agreement remained comparatively good (4.47°).
- Joint kinetics showed meaningful differences between systems, concentrated primarily during the swing phase of gait, with the ankle joint moment showing the largest discrepancy of the three joints assessed (RMSE = 0.20 Nm/kg), while joint power differences were most pronounced at the hip during the stance phase and smallest at the ankle.
- The markerless system tended to estimate slightly lower flexor moments than the marker-based system during the stance phase, and showed a specific difficulty capturing the brief period of ankle power absorption that marker-based data showed occurring at initial contact, though values from the two systems converged and were nearly identical for much of the remaining gait cycle at the ankle, despite a statistically significant difference being detected there.
What This Means for Movement Science Researchers and Movement-Recovery Assessment Teams
For researchers and teams assessing gait in children with cerebral palsy, this study offers a clear, specific picture of where markerless motion capture currently stands: sagittal-plane kinematics, and knee flexion in particular, showed strong agreement with marker-based motion capture even in this population with pronounced, varied gait deviations, supporting its practical use for these specific measures. At the same time, frontal- and especially transverse-plane kinematics, along with joint kinetics more broadly, showed larger and more inconsistent differences, suggesting markerless motion capture in its current form may be best suited to screening, longitudinal monitoring, and ecologically valid, real-world assessment settings in this population, rather than the kind of precise biomechanical modeling that surgical planning or detailed intervention decisions typically require. The authors' broader point, that marker-based systems themselves are not a true gold standard given their own known limitations, is a useful framing for interpreting any markerless-versus-marker-based comparison in a population where structural deformity complicates both approaches.
The Role of Theia3D in Movement Analysis
Theia3D was the markerless motion capture system evaluated in this study, compared directly against a 36-marker reference system for both joint kinematics and joint kinetics in children with cerebral palsy. This study's inclusion of joint kinetics specifically extends the evidence base beyond kinematic agreement alone, addressing a gap the authors note has received limited prior attention in movement-affected pediatric populations, where accurate joint moment and power estimation plays a central role in understanding muscle function and informing movement-related decisions.
Study Limitations and Future Directions
The authors note that differences in the underlying biomechanical models used by the two systems present an inherent challenge to direct comparability: the marker-based system's marker set allows 6 degrees of freedom at each body segment, while the markerless system's default model allows 6 degrees of freedom at the pelvis and 9 at each leg, a modeling difference the authors identify as a limitation of the study, while also noting the marker-based marker set offers specific benefits for assessing children with cerebral palsy given its detailed anatomical specification. The sample's gait pattern heterogeneity, with participants presenting a range of distinct gait deviations, increased variability in the kinematic and kinetic comparisons and may have reduced the statistical power available to detect differences at an individual level. The authors note that this study used the 2023 version of Theia3D, which is not the most current release, and that subsequent software improvements may affect how well these specific findings generalize to newer versions. They also point to differences in joint center and center-of-mass estimation between the two systems as a likely contributor to the kinetic discrepancies observed, identifying continued refinement of these specific estimations as necessary before markerless joint kinetics can be considered comparably accurate to marker-based systems. The authors identify larger studies involving broader movement-affected pediatric populations, and further validation specifically addressing joint kinetics, as important next steps before markerless motion capture can be more confidently applied to detailed intervention planning in this population.
Citation: Costa, R., Veloso, A., & Joao, F. (2026). Comparison of marker-based and markerless motion capture systems to assess gait kinematics and kinetics in children with cerebral palsy. Gait & Posture, 128, 110213. https://doi.org/10.1016/j.gaitpost.2026.110213
Link: Read the full study
Does Markerless Motion Capture Hold Up for Kinematics and Kinetics in Cerebral Palsy Gait?
This study compares Theia3D against marker-based motion capture for both joint kinematics and joint kinetics in children with cerebral palsy, finding strong sagittal-plane kinematic agreement, particularly for knee flexion, alongside larger differences in other movement planes and in joint kinetics. If your work involves gait assessment in populations with movement-related differences and you want to understand where markerless capture currently stands for kinetics as well as kinematics, talk to our team about applying Theia3D to your own research.