Enhancing the connectivity in a 2-node motor network using reinforcement learning-based closed-loop rTMS

Abstract

Abstract TMS has been extensively used to investigate the physiology of the central nervous system in health and disease. However, there is a strong need for a personalized, multi-locus, real-time stimulation procedure, which can be adaptively fine-tuned based on case-specific feedback. Here, we modulate the effective connectivity of the 2-node brain network from supplementary motor area (SMA) to primary motor cortex (M1) by closed-loop stimulation, optimized by application of an online reinforcement learning algorithm. This algorithm learns to identify the individually optimal phase of the ongoing μ-rhythm to be targeted by paired SMA-M1 TMS for maximized long-term enhancement of facilitatory effective connectivity between SMA and M1. This is one of the first demonstrations of true closed-loop stimulation, a crucially important step towards individualized highly-effective brain stimulation for therapeutic modulation of dysfunctional brain networks, e.g., the deficient SMA-M1 connection in motor stroke. Research Category and Technology and Methods Translational Research: 7. Responsive (Closed-Loop) Stimulation Keywords: EEG-TMS, Machine Learning, Closed-loop stimulation, Motor cortex