The Adaptive Cognitive Prosthetic: Developing a Closed-loop Micro-stimulation System to Produce Specific Neural Firing Patterns

Abstract

Brain-computer interfaces (BCIs) are clinically used to replace lost sensory or motor function. A neural prosthetic to replace lost cognitive function in patients with existing tissue damage, however, has yet to be developed. The Adaptive Cognitive Prosthetic (ACP), a closed-loop bidirectional BCI, was designed to replace the computational function of any brain region. Here, we test the ability of the ACP with its integrated adaptive algorithm to learn how to evoke behaviorally meaningful patterns of neural activity in mouse primary visual cortex with no pre-existing knowledge of the relevant stimulus-response function. First, we show that a novel operant conditioning paradigm teaches mice to discriminate between visual stimuli. We then characterize how effective natural visual stimuli are at eliciting specific and distinct neural firing patterns. Next, we find that the ACP can learn to direct neural activity towards a desired firing pattern by augmenting its electrical micro-stimulation pattern. Finally, we demonstrate that the effects of ACP micro-stimulation on neural activity overlap with the effects of natural visual stimulation on neural activity in PC space and, to an extent, in real neural response space. In sum, these results suggest that this closed-loop electrical micro-stimulation system may be able to learn stimulation patterns that reproduce specific, sensory-evoked neural responses. These results form a basis for eventually implementing the ACP in a clinical context.