A cognitive brain-computer interface prototype for the continuous monitoring of visual working memory load

We investigate the neural correlates of visual working memory using electroencephalography (EEG). Our objective is to develop a cognitive Brain-Computer Interface (BCI) able to monitor visual working memory load in real-time. A system with these properties would eventually have different applications, such as training, rehabilitation, or safety while operating dangerous machinery. The BCI performances were evaluated using cross-validation. With an appropriately chosen classification threshold, it is possible to detect high working memory load with a sensitivity of 68% and a specificity of 72%. However, it is well known that some subjects are BCI illiterate, meaning that up to 30% of the users have too high signal variability to use EEG-based BCI systems. If we analyse each subject individually, it is possible to detect high working memory load with a sensitivity of 78% and a specificity of 81% (accuracy = 81%) for a typical good subject. Changes due to visual working memory load were observed in frontal, parietal, and occipital regions.