Exploring transtibial prosthetic coupling (Part 1): Utilising adjustable sockets to influence prosthetic interface stiffness when comparing knee axis estimation methods.

Baldock, Michael; Harthikote Nagaraja, Vikranth; Dickinson, Alex; Curtin, Samantha

Abstract

BACKGROUND
Motion capture (MoCap) is widely used to study human ambulation. While optimisation to estimate the knee axis from MoCap markers has been explored in healthy populations [1-3], its effectiveness in transtibial prosthesis users remains unclear due to the increased marker artefact generated from prosthetic coupling movements. Adjustable sockets are gaining popularity in research and industry [4]. Their adjustments influence the ‘pseudo-joint’ interface properties; however, the changes to interface stiffness and coupling are unknown.

AIM
To utilise the variable interface properties of adjustable sockets to investigate knee marker optimisation strategy influence on prosthesis movement measurements.

METHOD
Gait was recorded for a unilateral transtibial participant across two sessions using MoCap for ten walking bouts at five socket tightness’s, providing ten interface conditions. Gold standard knee-axis estimation [1] was conducted on knee flexion data with the prosthesis doffed to create a baseline. Three functional methods [1-3] were then conducted on gait data, with socket-limb coupling movements calculated in four degrees of freedom (DoFs), with an estimated adduction movement5. Method accuracy was analysed against the baseline using root-mean-square-error (RMSE) of marker locations [1], pairwise RMSE and cross-correlations of mean-centred coupling results, and Spearman’s correlation to analyse within-session coupling range trends.

RESULTS
Sphere fitting with iterative noise reduction [2] proved to be the most accurate method of knee axis calculation from transtibial walking gait data throughout all tests. Principal component analysis (PCA) performed the worst throughout, probably because it optimises knee axis orientation but not location, as it retains proximal-distal marker placement error. However, the correlation of within-session trends between interface conditions was very strong. The axis location RMSE was comparable to virtual tests [1] which used smaller ranges of motion than gait (5° and 70°, respectively). The RMSE was largest in the transverse rotation measures for all methods; however, this DoF is known to be least accurate in MoCap.

DISCUSSION AND CONCLUSION
Two methods produced pairwise S/I RMSE of 3.3-3.6mm, which equates to 15% of pistoning displacement (22.3mm) for this user. Though limited to a single participant, the above results suggest that the sphere fitting algorithm with noise reduction should be used for knee joint axis estimation in transtibial gait analysis. Future work includes verification that the above stands true for more participants with differing socket, residuum, and gait characteristics and use in secondary data analysis to increase knowledge around prosthesis coupling trends.

REFERENCES
1.Ehrig, R.M. et. al. (2007). J. Biomech.
2.Halvorsen, K. et. al. (2003). J. Biomech.
3.Jensen, E. et. al. (2016). J. Biomech.
4.Baldock, M. et. al. (2023). J Neuroeng Rehabil.
5.Baldock, M. et. al. (2023). Gait Posture.

• Presentation Conference Type: Presentation / Talk
• Conference Name: ISPO World Congress 2025
• Start Date: Jun 16, 2025
• End Date: Jun 19, 2025
• Acceptance Date: Jan 20, 2025
• Deposit Date: Jun 27, 2025
• Peer Reviewed: Peer Reviewed