Loss of motor function after stroke substantially impacts the day-to-day life. In addition, the rehabilitation process is demanding and can require difficult physical exercises. Motor Imagery (MI), a technique where the user imagines the sensation of performing a movement, has been shown to activate similar brain activity as physically performing the movement. Brain Computer Interfaces (BCIs) with MI-based neurofeedback is a promising approach for rehabilitating loss of motor function without the use of physical exercise. For this technology to be used in rehabilitation, it is crucial that the neurofeedback reflects underlying neural activity that is functionally relevant for motor recovery. To this end, I believe personalisation is key. Therefore, in this thesis I have investigated important components for personalisation of a MI-BCI system for post-stroke rehabilitation. Specifically, I have 1) explored a distance-to-bound approach for adapting the BCI task difficulty, 2) investigated the impact of continual visual feedback on MI-related cortical activity, and 3) analysed the feasibility of extracting a novel motor-related feature.
I suggest that adjusting the difficulty of an MI-BCI, using distance-to-bound, targets stronger Event Related Desynchronisation (ERD) during neurofeedback training. However, strong ERD does not directly correlate with improved motor function, highlighting the importance for personalisation. I further provide evidence, in a group of stroke patients, that continual visual feedback does not interfere with the MI-related cortical activity, nor does it lead to stronger activity. Finally, I show that transient heterogenous features in the beta band, so called beta bursts can be extracted from both the lesioned and healthy hemispheres of stroke patients. These results contribute both methodologically and scientifically in building a stronger foundation of knowledge for the development of MI-BCI rehabilitation after stroke.
However, there is still need for more research to fully understand the complexity of neurofeedback stroke rehabilitation. I believe this thesis sheds light on important considerations when developing the different components of a BCI designed to promote motor recovery after stroke.