Communication is fundamental to human interaction, yet neuromuscular disorders can severely restrict it. For individuals with advanced conditions such as Locked-in syndrome, maintaining reliable communication is crucial, and Brain-Computer Interfaces (BCIs) offer a promising approach. This study presents a movement-related cortical potential (MRCP) based speller designed for users with severe motor impairments. By detecting brain signals elicited by self-initiated movements that function as a brain click, the system enables intuitive control across various interfaces and shows improved performance when the classifier is adapted to specific applications. To identify the most robust neural signals for control, we evaluated five right-hand gestures: Fist, Pincer, Y, Pistol, and Hand-up, performed solely to evoke movement-related signals serving as the brain-click input. A two-stage training strategy was employed. First, the classifier was trained on generalized cue-based data aligned to visual cues. Then, as a new contribution, it was retrained using data recorded during real speller operation, allowing adaptation to the user's online brain-click patterns and substantially improved practical performance. Across participants, retraining increased average true positive rate from 27.3 percent (1.1 false positives per minute (FP/min), 1.1 characters per minute (CPM)) to 63.0 percent (1.0 FP/min, 2.7 CPM), highlighting the benefit of task-specific adaptation. While no statistically significant differences were observed across gestures, Hand-up consistently yielded the highest detection accuracy and was selected most frequently, suggesting its suitability as a default control signal. Compared to earlier MRCP based systems that rely on overt movement, our cue-aligned approach achieved higher true positive and lower false-positive rates when adapted to the specific interface, representing an important improvement toward more efficient communication for users with severe motor impairments. While not tested in home settings, these results point toward the possibility of a home-usable MRCP speller and highlight the value of personalized, adaptive BCI control. .
Suwandjieff et al. (Mon,) studied this question.