Gait kinematics and kinetics of children with idiopathic toe walking: Insights from statistical physics - Escuela Universitaria de Fisioterapia de la ONCE
Escuela Universitaria de Fisioterapia de la ONCE - WELEUF_PUBLICACIONES_DETALLE
Gait kinematics and kinetics of children with idiopathic toe walking: Insights from statistical physics
Resumen
Introduction
Idiopathic toe walking (ITW) is a heterogeneous form of gait with limited or absent heel strike at the beginning of the cycle, and its diagnosis is of exclusion of other neurological or musculoskeletal conditions. Particularly difficult and relevant is to early discriminate mild bilateral spastic cerebral palsy (BSCP) and genetic forms of spastic paraplegia from ITW [1]. In the context of a larger project intended to understand the gait differences between BSCP and ITW, we have first tested the power of several metrics from statistical physics (entropy and irreversibility) which assess gait signals in terms of information on embedded stochastic processes (which are supposedly reflecting the quality of neural control), and of deep learning [2] to classify healthy and ITW gait kinetic and kinematic time series.
Research question
Is the temporal and causal structure of gait signals altered in children with ITW, who are free from brain damage?
Methods
We evaluated 96 healthy children (mean age: 9 years, 50% male) and 58 children with ITW (9 years, 76% male) with clinical examination and 3D gait analysis (545 vs 265 gait cycles, respectively).
We extracted gait kinematic (pelvis, hip, knee, ankle, and foot angles in three planes) and kinetic (ground reaction force in the three coordinates and hip, knee, and ankle moments and powers) normalized time series. Entropy and irreversibility indices [[3], [4], [5]] were calculated from each gait
Results
Entropy indices clearly distinguish ITW from healthy cycles (85% of precision for kinematic time series), and so do irreversibility indices (73% of discriminatory power for moments and powers). Deep learning shows only a minor advantage in terms of classification performance (ankle flexion moment CS: 91.9%, forefoot protonation CS: 91.6%).
Discussion
Statistical physics metrics find abnormalities in the organization of ITW gait signals that distinguish them from healthy, heel-strikers, in a way similar to a deep learning algorithm, which is expected to avoid decisions regarding previous feature selection.