Artificial Intelligence platforms for translatable closed-loop neuromodulation

Abstract

Closed-loop neuromodulation and direct neural interface systems provide powerful tools for elucidating the causal effects of regulating neurophysiological processes and restoring function, and promise to enable precision therapies for a wide range of diseases. However, the complexity of controlling the neural dynamics and designing optimal closed-loop control systems for neural interfaces with often large parameter spaces is a major barrier. Furthermore, implementing advanced analytic workflows for real-time processing of large-scale neural data can be prohibitive for translating closed-loop control systems into experimental and clinical settings. Recent advances in Artificial Intelligence (AI) have created unprecedented opportunities to address these problems. I will discuss our ongoing efforts in leveraging AI technologies to develop platforms for designing, prototyping and implementing 1- intelligent Closed-Loop Neuromodulation systems that can automatically learn optimal neuromodulation control policies from interacting with the nervous system, 2- interpretable AI pipelines for automated biomarker discovery and quantifying the effects of neuromodulation, and 3- hybrid edge-cloud computing infrastructures for distributed implementation and dissemination of AI pipelines and workflows. We are developing these platforms in the context of neuromodulation for memory enhancement, deep brain stimulation for movement disorders, epilepsy, and peripheral neuromodulation for restoring organ functions using an array of experimental and computational modeling approaches.