| The long-term goal motivating this research is to design a Neurorobotic Interface that will allow the restoration of lower limb control after a neurologically complete spinal injury. The project's central hypothesis is that populations of single neurons in the hindlimb representation of the motor cortex could (1) be shown to encode specific kinematics of a skilled hindlimb movement, and (2) continue to encode for movement intent following a complete spinal transection. Adult rats were trained to press a lever with a hindlimb in response to an audible cue. Following behavioral training, animals were implanted stereotaxically with chronic indwelling arrays of 50 mum stainless steel microwires into the hindlimb representation of the motor cortex. Spiking activity was then recorded from ensembles of single neurons while the animals resumed the skilled hindlimb lever press behavior. Offline analysis of spiking activity, with lever position data, quantified the degree of correlation between neuronal firing and the kinematic parameters of the movement, as well as the temporal tuning properties of the cells. Later, control of reward delivery was changed from lever-press activity to the value of a weighted-sum average of the activity of the neural ensemble, calculated in real time during the experiments. Once neural control had been established, the lever was removed, and finally a complete mid-thoracic spinal transection surgery was performed. Offline analysis of the population activity showed that cortical firing patterns could encode for the intention to move, with or without actual limb movement. Decoding accuracy changed with algorithm update frequency, possibly indicating that functional reorganization took place in the cortex because of the neurorobotic paradigm. Following transection, there were decreases in both the proportion of cells modulating their activity in response to the audible cue, and in the firing rates of those cells. Decoding accuracy did not significantly decrease, however, as a result of spinal transection, and the significant effect on overall accuracy by algorithm update frequency was not apparent following transection. These results indicates that the presence or absence of an injury condition does not irrevocably prevent the use of a Neurorobotic Interface, as movement intent is preserved following injury. |
