3D body centre of mass trajectory in locomotion: comparison between different measurements methods

Introduction and Objectives: Body centre of mass (BCoM) trajectory is largely used to describe human and animal locomotion with the final aim to estimate the mechanical work performed by muscles. The BCoM can be calculated through 2 different methods: i) by double integration of ground reaction forces (GRF) and ii) by the weighted mean of the local centres of mass of the body anatomical segments, which are collected through an optoelectronic system. Although both methods are currently exploited, the use of the GRF is considered the gold standard as it provides a (relative) BCoM trajectory as caused by both hard and soft tissue motion. However, it allows capturing only a limited number of strides and repeated trials are requested for reliable results. Kinematic measurements, conversely, can be carried out while moving on a treadmill, so that an adequate strides number and speed could be collected. Several biomechanical models are available to analyse the kinetics of human body and the use of different anthropometric tables could bring discrepancies in the BCoM computation. In addition, markers placement, skin motion and the rigid segment assumption could lead to systematic errors. Previous studies compared these methods, but only few locomotion paradigms were investigated, mainly analysing only BCoM vertical motion. Aim of this study was to compare the 3D BCoM trajectory and the associated mechanical work (WEXT) concurrently calculated by means of GRF and kinematic in 5 gaits and a wide range of speeds. Kinematic results were compared across 5 different sets of markers in order to verify their reliability.
Methods: One participant (1.78m, 63kg), performed different gaits on treadmill at incremental speeds: walking from 0.28 to 1.94m/s (increment 0.28m/s); running from 2.22 to 5.56m/s (increment 0.28m/s); race walking from 2.22 to 4.17m/s (increment 0.28m/s), mono- and bi-lateral skipping from 0.83 to 3.06m/s (increment 0.56m/s). An 8-camera Vicon system (Oxford Metrics, UK) sampled kinematics data at 300Hz for 1 minute at each velocity. Simultaneously a Mercury LT med treadmill (HP Cosmos, Germany), equipped with 4 3D strain-gauge force traducers, recorded GRF at 900Hz.
5 markers set were compared at the same time: 1) a single marker placed on the 7th cervical vertebra (C7); 2) the mean of anterior and posterior superior iliac spines (Spinae); 3) a 11-segments body model based on Dempster tables (18mkr); 4) a 14-segments body model based on De Leva tables (DeLeva); 5) a 14-segments body model based on the Vicon Plug-in-Gait model (PIG). In addition BCoM was obtained by double integration of the filtered GRF (low pass filter 4th order Butterworth with 30Hz cut off frequency).
BCoM trajectory of each stride (local coordinates on treadmill) was isolated and the loop forced to close, according to Minetti’s method [1]. A point-by-point 3D root mean square (3D RMS) was computed between the (reference) trajectory evaluated with GRF method and the one calculated with each of the 5 kinematic models. WEXT, i.e. the positive work done to accelerate and raise BCoM, was calculated for each trajectory by summing the increment of total energy time course.
Results: The 3D RMS between the trajectories calculated with GRF method and kinematic methods indicated that the marker set exploiting C7 and Spinae showed the greatest discrepancy from GRF, whereas 18mkr and DeLeva models well matched all gaits and PIG was in accordance to GRF method only in running (Fig 1). Although the BCoM trajectory in race walking showed a small discrepancy in terms of RMS, its 3D contour was very different when compared to the one computed from GRF. WEXT obtained from kinematic showed the slightest discrepancies when adopting PIG, 18mkr and DeLeva models (Table.).