Transcranial magnetic stimulation-based brain-computer interface with neurofeedback facilitates motor imagery of different daily life hand actions

Background: Non-invasive brain-computer interfaces (BCIs) in combination with motor imagery (MI) and neurofeedback (NF) enable users to modulate motor-related brain activity without producing actual motor output. While most non-invasive MI-based BCIs decode only gross movements, daily tasks require finer finger and hand control. We developed a novel transcranial magnetic stimulation (TMS)-based BCI training using MI and NF to reinforce representations of complex hand actions in the brain. This proof-of-concept study examines the feasibility of using this BCI for hand function training. Methods: Twelve healthy adults underwent one motor execution (ME) and three TMS-based BCI training sessions on three hand actions (holding a bottle, turning a key, and opening the hand). A personalized, adaptive support vector machine (SVM) ensemble predicted the hand action during MI trials using motor evoked potential (MEP) patterns, providing feedback based on the classification outcome. Block-wise balanced decoding accuracy, derived from an SVM classifier with leave-one-trial-out cross-validation, was chosen as the outcome measure and analyzed using linear mixed-effects models to investigate training and learning effects. Results: With three sessions of TMS-based BCI training using MI and NF, the three distinct hand actions could be decoded from MEPs and the decoding accuracy was significantly above the empirical chance level. Participants could modulate their brain activity when NF was present across sessions, but this did not transfer to a condition without NF. Additionally, to evaluate different classifier training approaches, we found that both ipsilateral and contralateral ME-trained classifiers could decode imagined hand actions above chance level. Furthermore, training the classifier on ME data from the same participant was significantly better than using data from other participants for the training. Conclusions: We developed and tested a novel, personalized, and adaptive TMS-based BCI for complex hand actions. Our findings suggest that healthy adults can modulate brain activity during complex hand actions with the guidance of NF. These results underscore the translational potential of TMS-based BCI for neurorehabilitation, providing a pathway for MI training of functional hand actions in individuals with severe unilateral impairments who can be guided by reference signals extracted from movements with the less-affected hand. Trial registration: This study was registered at Open Science Framework (https://doi.org/10.17605/OSF.IO/84HUJ).