UC Davis

Despite advances in our understanding and management of spinal cord injuries (SCI), less than 1% of individuals completely recover by hospital discharge. Restoration of voluntary motor function remains one of the most desired therapeutic outcomes for individuals with these injuries. Brain-machine interfaces (BMI) have been shown to decode neural signals about movement intention and use these signals to control an external device (e.g., a cursor). Since neuronal circuitry remains largely preserved in the spinal cord distal to the injury, using decoded neural signals to control stimulation of the spinal cord itself could allow paralyzed individuals to produce self-directed movements. While the majority of lower limb BMI research has focused on restoring rhythmic locomotor movement patterns and has garnered success as it translates to human clinical trials and device development, all of these approaches have failed to generate completely independent movement in part due to an ability to restore postural stability. Compromised postural control notably increases risks of falls, carrying additional personal physical and financial burden. Therefore, it is necessary to develop a BMI to restore this critical aspect of lower limb function after SCI.

The long-term goal of this work is to develop neuroprosthetic interventions to improve motor function and improve quality of life for individuals with debilitating neurological conditions. The aims of this project specifically were to understand (1) how the cortex – the most ubiquitously used signals in neuroprosthetics – encodes for postural control, (2) the extent to which this information can be decoded to inform BMI design, and, most importantly, (3) whether or not this remains true after traumatic spinal cord injury. We demonstrate that neurons across the sensory and motor cortex convey significant information about postural perturbations, even after two different models of spinal cord injury (moderate contusion and complete transection). This information is encoded in parallel streams by speed-scaling and direction-dependent neural responses. Ground reaction forces, notably those most altered by mediolateral shifts in center of pressure, can be decoded based on the latent dynamics of the motor cortex, with different perturbation directions requiring unique computations that can be scaled with perturbation speed. After injury, the responsiveness of individual neurons to perturbations and the information conveyed by the neurons that respond is attenuated but remains significantly above chance (and can be further enhanced with physical rehabilitation); however, as a population, cortical dynamics are largely preserved. Thus, the cortex, which is already a target in multiple brain-machine interface trials, should be considered as a control signal for a postural neuroprosthetic.