Developing real-time personalized TMS to target residual corticospinal connections after stroke

About This Grant

PROJECT SUMMARY Stroke commonly disrupts the corticospinal tract (CST) and impairs hand function. Transcranial magnetic stimulation (TMS) interventions that target and strengthen residual CST connections are promising candidates for improving poststroke hand function. To maximize their therapeutic effects, such interventions must repeatedly activate the residual CST and enhance its neural transmission. We and others recently showed in neurotypical adults that resting brain activity spontaneously alternates between EEG activity patterns (brain states) that predict strong and weak CST activation. TMS interventions also preferentially enhance CST transmission when delivered during strong CST states but instead diminish CST transmission when delivered during weak CST states. However, virtually all poststroke TMS interventions are uncoupled from the current brain state, such that only a fraction of TMS stimuli coincide with brain states during which the beneficial effects of TMS are likely to be strongest. To resolve this issue, poststroke TMS interventions should be delivered solely during brain states reflecting strong CST responses. Given that each stroke survivor has a unique pattern of brain damage and recovery-related brain reorganization, these brain states must be fully personalized. We recently developed a personalized machine learning framework that successfully identifies electroencephalography (EEG) activity patterns that predict strong and weak CST states in neurotypical adults. Our framework is fully personalized and is therefore unaffected by lesion-related changes in brain structure and/or function, making it ideal for application in the poststroke brain. In this project, we will use this framework to establish the mechanistic rationale and methodological foundation for future personalized brain state-dependent TMS interventions that target and strengthen the residual CST after stroke. In Aim 1, we will use our machine learning framework to identify personalized brain states that predict strong and weak residual CST activation in chronic stroke survivors; we will also evaluate relationships between our framework’s performance and functional and structural metrics of poststroke CST pathway integrity. Results of Aim 1 will establish poststroke brain state-dependency of residual CST output and the relationship of this state-dependency to CST integrity. In Aim 2, we will develop and validate a real-time EEG algorithm that accurately delivers TMS during personalized brain states reflecting strong and weak CST activation in neurotypical adults. Results from Aim 2 will demonstrate the technical feasibility of personalized, real-time brain state-dependent TMS. Overall, this project fits the scope of the NIMH/NINDS R21 mechanism because it will develop a novel neuroengineering approach that can in the future enhance residual CST transmission and promote paretic hand function in stroke survivors.