A computational framework to study neural-structural interactions in human walking

Abstract

Neuroscientists researching locomotion take a top-down approach by elucidating high- level cortical control circuits. In contrast, biomechanists prefer to focus on structural and mechanical aspects of the legged movement apparatus. We posit that studying interplay between neural co-ordination and legged biomechanics can yield crucial insight into (a) motor control and (b) human leg morphology. Physiological facts indicate that muscle actuator state (activation, length and velocity) is key to this neural-structural interplay. Here we present a novel model-based framework to resolve individual muscle state and describe neural-structural interactions in normal gait. We solve the inverse problem of using kinematic, kinetic and electro-myographic data recorded on healthy humans during level-ground,self-selected speed walking to estimate state of three major ankle muscles. Our approach comprises of two steps. First, we estimate neurally-controlled muscle activity from EMG data by building on statistical and mechanistic methods in the literature. Second, we perform a system ID on a mechanistic (Hill-type) model of the three muscles to nd tendon morpho- logical parameters governing evolution of muscle length and velocity. We implement the parameter identication as an optimization based on the hypothesis that neural control and lower limb morphology have co-evolved for optimal metabolic economy of natural walking.

(cont.) We cross-validate our framework against independent datasets, and nd good model-empirical ankle torque agreement (R 2 = 0.96). The resulting muscle length and velocity predictions are consistent with in vivo ultra- sound scan measures. Further, model predictions reveal how leg structure and neural control come together to (a) dene roles of individual plantar exor muscles and (b) boost their joint performance. We nd that the Soleus operates as a steady ecient force source, while the Gastrocnemius functions as a burst mechanical power source. An analysis of the estimated states and optimized parameters reveals that the plantar exors operate jointly at a net mechanical eciency of 0.69 ±0.12. This is roughly three times higher than the maximal eciency of skeletal muscle performing positive work. Our results suggest that neural control may be tuned to exploit the elasticity of tendinous structures in the leg and achieve the high walking economy of humans.