実歩行計測データからの歩行神経回路網の推定

抄録

Locomotion can be spontaneously generated in the relatively lower nervous system by autonomously coordinating the rhythmic activity of a Central Pattern Generator with afferent sensory information from the proprioceptors; it is not precisely controlled by the higher center. Inspired by these physiological findings, some computer simulation models of human locomotion have been proposed that mimic human locomotory neuronal mechanisms. These models, however, do not correspond neuro-physiologically to the actual human nervous system, and simulating successful locomotion does not necessarily imply elucidation of the neuronal mechanism of human locomotion. In this study, in contrast to these synthetic approaches, we attempted to estimate the human locomotory neural network, based on an actually measured kinematic and kinetic data set for human bipedal locomotion. The kinematic and kinetic data may describe structure and behavior of the neural network, since human motion is an output of the nervous system, as well as an input to it, in as much as the resultant motion is perceived by proprioceptors. Using the measured data for natural human walking and a two-dimensional musculo-skeletal model of a human lower extremity, muscular forces while walking were calculated by solving the inverse dynamic problem. The motor commands sent to muscles by alpha motoneurons can be estimated based on the calculated muscle forces. Two kinds of proprioceptors (the Golgi tendon organ and the muscle spindle), and tactile receptors on the plantar surface of the foot were considered. Activities of these sensors can also be estimated from changes in the physical quantities (such as muscle strain) that are actually perceived by each of these sensors during locomotion. The signal generated by the Central Pattern Generator is modeled as a sine curve, having a period equal to the walking period, since it basically generates the simple flexion-extension motion of a limb while walking. The estimated motor commands of the alpha motoneurons were then reconstructed by summation of the estimated signals of the proprioceptors, the foot tactile receptor, and the rhythm generators, according to the basic structure of the peripheral nervous system, such as reciprocal innervations. The calculated result suggests that the activity of alpha motoneurons can be represented by the weighted linear summation of the sensory and rhythmic inputs, indicating that locomotion can be generated autonomously by integration of the sensory inputs according to the structure of the peripheral nervous system. The estimated weights of connections were then compared to select important sensory information for generating locomotion. As a result, it was found that locomotion could be generated by a relatively simple neural network, and the sensory inputs, especially the GII signals of the muscle spindles, are more dominant than the rhythm pattern generator for generation of steady state locomotion.