Variations of lower-limb joint kinematics associated with the use of different ankle joint models

Skeletal computational models relying on global optimisation are widely used alongside gait analysis for the estimate of joint kinematics, but the degrees of freedom (DOFs) and axes definitions to model the ankle complex are still debated. The aim of this paper is to establish whether ankle modelling choices would also critically affect the estimate of the other joints' kinematics. Gait and MRI data from fifteen juvenile participants were used to implement three ankle joint models (M1, one-DOF sagittal motion; M2, two-DOFs sagittal and frontal motions; M3, three-DOFs) as part of a full lower-limb skeletal model. Differences in lower-limb joint and foot progression angles calculated using global optimisation were evaluated both at individual and group level. Furthermore, the influence of these differences on the correlations between joints and on the calculations of the root mean square deviation (RMSD) were investigated. Inter-model variations at individual level reached up to 4.2°, 9.1°, and 15.0° for hip flexion, adduction, and rotation, respectively, and up to 6.5° for knee flexion. Despite the tibiotalar axis being the same for all models, up to 19.3° (9.1° on average) larger dorsiflexion was found at push-off with M2. A stronger correlation between foot progression and ankle and knee sagittal movements was found for M1. Finally, RMSD led to inconsistent ranking of the participants when using different models. In conclusion, the choice of the ankle joint model affects the estimates of proximal lower-limb joint kinematics, which should discourage comparisons across datasets built with different models.