CLOsing the loOP on non-invasive brain sTiMulation with deep neural networkS (LOOP-TMS)

ongoing

We aim to develop and validate adaptive closed-loop brain stimulation using advanced real-time electroencephalography-transcranial magnetic stimulation (EEG–TMS) technology guided by probabilistic deep artificial neural networks. TMS is performed routinely worldwide in tens of thousands of healthcare institutions for treating a wide variety of brain disorders. However, current TMS protocols are limited by small therapeutic effect sizes and large portions of non-responders. A significant problem is that one size does not fit all: existing TMS protocols are non-personalized, operate without information of instantaneous brain states, function in an open-loop manner, and lack adaptability. To address this, we have developed brain state-dependent stimulation in the ongoing ERC Synergy project ConnectToBrain. Real-time EEG–TMS has enabled reliable prediction of high- vs. low-excitability brain states and coupling of TMS to a prespecified state. While this is major progress, brain state-dependent stimulation is still open-loop and non-adaptive. The key innovation in our current proposal is implementing truly adaptive closed-loop algorithms that continuously monitor TMS effects and optimize stimulation parameters in real-time to maximize beneficial long-term changes in targeted brain networks. Recent advances in machine learning now make this sophisticated approach possible. This project represents a significant leap beyond ConnectToBrain by creating a user-independent, fully automated, adaptive closed-loop real-time EEG-TMS system ready for further testing in clinical trials. Our technology has breakthrough potential as a paradigm-shifting innovation in non-invasive therapeutic brain stimulation, addressing an urgent societal and medical need for more effective, individualized treatment of brain disorders.