Kinetic errors in 3D gait analysis driven by inaccurate inertial parameter estimation of prosthetic limbs

The conventional gait model (CGM), which forms the basis for software implementations such as Vicon Nexus’ plugin gait model (PiGM), has been designed to model standard anatomy. When used to assess patients with lower limb prostheses the underlying modelling assumptions are challenged, introducing several inaccuracies [1]. One such inaccuracy is the method used by CGM implementations to estimate the proportional mass, location of the centre of mass and moment of inertia (MoI) of various body segments using regression equations [2]. Dempster developed these equations experimentally, measuring cadaveric limb segments [3]. Prosthetic limb segments and human limb segments are different, and therefore, inertial parameters estimated using these equations will not be correct, leading to errors in the kinetic data [2]. How different are prosthetic limb segment inertial parameters from intact limb segments, and how does this affect the kinetics output from 3D gait analysis? Using patient images and prosthesis prescriptions, Fusion360 (version 2.0.15050) computer aided design (CAD) software was used to model prosthetic body segments for 16 subjects, and estimate their inertial parameters. This method was validated by comparing CAD model inertial parameters to those measured using ‘gold standard’ practical experimental techniques – weighing scales, the reaction board method and the oscillation technique – of sample prostheses. PyCGM2 (version 4.3-beta) (an opensource python replication of the CGM) [4] was used to reprocess the subjects PiGM kinetics incorporating the CAD derived inertial parameters. These kinetics were compared with the standard PiGM kinetics using statistical parametric mapping in 1 dimension [5]. Shank mass and MoI were reduced by 32.5% (±25.7%) and 50.3% (±21.8%), and Foot mass and MoI were reduced by 30.3% (±18.7%) and 64.5% (±15.2%) (±SD), when comparing CAD estimates with Dempster regression equation estimates. Differences (p<=0.002) were found in both hip and knee sagittal plane kinetics during early and terminal swing phases, with peak moments, occurring at terminal swing, having been reduced by 37.3% (±22.0%) (at the hip) and 33.6% (±15.8%) (at the knee) using CAD derived inertial parameters.Download : Download high-res image (158KB)Download : Download full-size image Sagittal plane hip and knee kinetics during early and late swing are overestimated when 3D gait analysis, using the standard implementation of the CGM, is performed on patients with lower limb prostheses. Unsurprisingly the effect is mostly observed during swing phase when the ground reaction force is zero, leaving only the inertial terms. Swing phase moments are important for understanding the effort required to progress the prosthetic limb. Future research should develop regression equations for the approximation of prosthesis inertial parameters, which would offer quicker estimates, with similar accuracies, whilst being more repeatable than the CAD based protocol developed here.