Optimising Continuous Control of Real-Time Brain-Computer Interfaces Through Trial Length and Feedback Update Interval Selection

Brain-computer interfaces (BCIs) offer promising potential to aid neurorehabilitation by transforming motor imagery (MI) signals into control commands, bypassing damaged neural pathways to support motor recovery. However, a key challenge in BCI research is achieving an effective balance between classification accuracy and real-time responsiveness, as both are critical for enhancing user embodiment and control for neurorehabilitation outcomes. This study investigates the impact of trial length and feedback update interval (FUI) on classification accuracy in an MI-based BCI system. Using EEG data from five subjects across 50 sessions, we evaluated classification performance across various trial length (1-5 seconds) and FUI (0.2-1 second) configurations. Results revealed that both trial length and FUI significantly influenced classification accuracy, with longer trial length (4-5 seconds) and FUI (0.4-1 seconds) yielding the highest accuracy. However, post-hoc analyses indicated a saturation effect, with no significant differences in the accuracy for these parameters. These findings underscore the importance of balancing signal stability with responsiveness for optimal BCI performance, providing insights into parameter settings that can enhance BCI usability in neurorehabilitation. Future work may explore adaptive approaches to dynamically adjust these parameters based on real-time requirements, potentially offering a more responsive and efficient BCI for clinical rehabilitation.