Study at a glance

This study assessed the test-retest reliability and measurement error of Theia3D markerless gait kinematics and kinetics in 29 healthy older adults tested on two separate occasions, finding good overall reliability for hip, knee, and ankle joints, with lower reliability for pelvis angles.

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: v2023.1.0.310 (patch 1).

Study Overview

Design

A prospective, within-examiner test-retest study in which 29 healthy older adults underwent markerless gait analysis on two separate occasions, approximately nine days apart, with the same assessor conducting both sessions. Reliability and measurement error were calculated by comparing kinematic and kinetic data between the two sessions for each participant.

Participants

Twenty-nine healthy, independent, community-dwelling older adults (13 female, 16 male; mean age 75.2 ± 7.7 years; BMI 25.0 ± 2.6 kg/m²) able to walk 100 meters independently without walking aids, with no cognitive impairment (MoCA ≥ 23) and no clinical condition or medication affecting their ability to walk.

Data Collection

Gait data were collected using 8 synchronized Theia3D video cameras alongside 3 force plates, with participants walking back and forth across a 12-meter laboratory walkway at their normal comfortable pace, wearing their own everyday clothing and sports shoes, until approximately ten valid gait cycles were captured on each side. Testing was repeated in a second session roughly nine days later under the same conditions.

Analysis

Video data were processed through Theia3D to generate joint kinematics, with joint moments calculated via inverse dynamics in Visual3D. Eight gait cycles per session were averaged per participant. Test-retest reliability was assessed using pointwise and integrated intraclass correlation coefficients (ICC) across the full gait cycle and for discrete peak values, with measurement error quantified using standard error of measurement (SEM) and minimal detectable change (MDC).

Key Findings

  • Hip, knee, and ankle kinematics showed good overall test-retest reliability. Integrated ICCs ranged from 0.53 to 0.88 across all kinematic measures, with sagittal-plane ankle motion showing the most reliable results of any kinematic parameter, including excellent reliability (ICC > 0.90) across 60-80% of the gait cycle.
  • Pelvis angles and hip sagittal-plane motion during swing were the least reliable kinematic measures. Sagittal and transverse pelvic segmental angles fell below the 0.75 threshold for good reliability (integrated ICC of 0.53 and 0.72 respectively), and hip flexion/extension reliability dropped during part of the swing phase.
  • Measurement error for joint angles stayed low and practically usable for detecting real change. SEM values ranged from 1.0° to 2.2° across the gait cycle, with MDC (the threshold for a confidently real change) ranging from 2.9° to 6.1°, though SEM expressed as a percentage of each measure’s range was notably higher for frontal- and transverse-plane measures, especially the knee, reflecting their smaller overall range of motion.
  • Joint moment (kinetic) reliability was good to excellent for most measures, and compared favorably to prior marker-based reliability data. Integrated ICCs for kinetics ranged from 0.74 to 0.89, with the clearest exception being hip rotation moments in the transverse plane; the authors note their kinetic ICCs were generally higher than those reported in a prior marker-based reliability study in a similar population, despite markerless participants wearing their own unrestricted clothing.
  • Whole-cycle reliability analysis revealed patterns a single peak-value summary would have missed. For sagittal-plane ankle angle, for example, integrated ICC (0.86) and peak flexion ICC (0.97) diverged noticeably, and pointwise ICC dipped close to the 0.75 threshold during early-to-mid stance, a pattern the authors argue underscores the importance of examining reliability across the full gait cycle rather than relying only on summary or peak values.

What This Means for Researchers Working With Older Adults

For research teams tracking gait changes in older adults over time, whether monitoring age-related decline, disease progression, or response to an intervention, this study provides concrete MDC values that can be used to judge whether an observed change in someone’s hip, knee, or ankle kinematics or kinetics reflects a real difference rather than normal measurement variability. The one clear caution flagged by the authors is pelvic tilt and rotation, along with hip motion during swing phase, where lower reliability means more caution is warranted before interpreting a change as meaningful. Because the study allowed participants to wear their own everyday clothing rather than requiring minimal, fitted attire, the authors argue these reliability figures reflect the kind of liberal, low-friction setup that would be realistic in a real-world research or assessment environment.

The Role of Theia3D in Movement Analysis

Theia3D allowed the research team to assess gait reliability and measurement error across a full testing protocol without requiring markers, tight clothing, or the extended setup time that could be especially burdensome for older adult participants across two separate visits. The study is also notable for including joint kinetics (moments), not just joint angles, in its reliability analysis, filling a specific gap the authors note had not previously been reported for this markerless system.

Study Limitations and Future Directions

The authors note that no specific clothing instructions were given to participants, which could have affected pose estimation consistency between the two sessions, though they point out that prior research (Keller, V.T., Outerleys, J.B., Kanko, R.M., Laende, E.K., & Deluzio, K.J., “Clothing Condition Does Not Affect Meaningful Clinical Interpretation in Markerless Motion Capture,” Journal of Biomechanics, 2022) suggests clothing has only a minimal effect on clinically meaningful interpretation, and that allowing participants to wear their own clothing actually reflects one of markerless capture’s practical advantages for real-world research settings. Data collection took place in a laboratory with natural light restrictions that could affect video quality, although lighting conditions were kept similar between the two sessions. The authors also note their linear time-domain registration of gait cycle data could overlook temporal variations between strides, and suggest a non-linear registration approach could be explored in future work, though they don’t believe this significantly affected their results given the clearly identifiable gait events in healthy participants.

Citation: Carvalho, A., Vanrenterghem, J., Cabral, S., Assunção, A., Fernandes, R., Veloso, A.P., & Moniz-Pereira, V. (2024). Markerless three-dimensional gait analysis in healthy older adults: test-retest reliability and measurement error. Journal of Biomechanics, 174, 112280.

Link: https://doi.org/10.1016/j.jbiomech.2024.112280

How Confident Can You Be in the Gait Changes You’re Tracking?

Knowing whether a change in someone’s gait over time is real, rather than just measurement noise, is essential for any program tracking mobility in older adults. If you’re building a gait monitoring protocol for an aging population and want to understand what markerless capture can reliably detect, talk to our team about how Theia3D’s reliability profile fits your specific tracking needs.

Recent Posts