This research proposal aims to investigate the optimal design features of intracortical motor neuroprosthetics by analyzing: 1) The ability of specific algorithms to predict movements from neural activity, 2) the ability of non-human primate subjects to control neurally-derived signals in a behaviorally useful way, and 3) explore the abilities of humans to control computer cursors with neural activity. First, discrete and continuous movement models will be evaluated to determine the minimum amount of data necessary to build robust decoders. This relates to clinical calibration: How much time will be required for patients to build decoding models that they can then use? Next, I will evaluate if and how animals can control a computer cursor whose position is driven by decoded neural activity alone: Whether they can generate stationary holds, ballistic reaches, and complex trajectories. Finally, I will test the decoding system in humans by participating in ongoing human multi-electrode recording. I hope to implement a real-time neural activity decoding system to test whether human subjects can rapidly acquire useful, accurate control over a movement signal. These investigations will shed light on which parameters are useful for achieving such control and yield insight on the optimal balance between computer and human learning in the closed-loop control system to restore maximum functional independence.
