Gait pattern prediction via bilateral neural ensemble recordings in motor cortex in rats.

We are interested in developing and building an experimental system for controlling a quadruped robot with a brain-derived signal of a rat. We recorded large ensembles of neurones in left M1 and right M1 and S1 using chronically implanted multi-electrodes arrays (up to 96 electrodes) in rats. The interpretation of such signals has proceeded to an advanced stage in animal experiments. Recently, this highly documented and efficient technology has given many concrete results and has opened new perspectives in the neuroprosthetics field. In our project we correlated motor area signals with the speed of a rat when walking on a treadmill. In parallel, we have performed an accurate analysis of rats kinematics. We performed the gait analysis thanks to an ad hoc motion capture system. We are able to reconstruct a 19 DOF (degree of freedom) simulated model of the rat skeleton for different speeds and different gaits patterns (walk to trot). We can now predict any of the 19 joint angles positions given brain derived signal. Preliminary results of our models suggest that with a high dimensional signal from cortical activity we can predict the actual gait patterns well (up to R2 = 0.64).