A Longitudinal Investigation of Non-Invasive P300-Based Brain-Computer Interface

Brain-Computer Interface is an alternative method of communication. Amyotrophic lateral sclerosis (ALS) patients can derive benefit from BCI use, as ALS can result in locked-in syndrome. The present BCI operates via event- related potentials (ERPs) extracted from the EEG. Items (i.e., alphanumeric characters and keyboard commands) attended to by the subject should produce a P300 ERP; unattended items should not. Stepwise linear discriminant analysis is used to classify the items as attended or unattended. Many studies show that ALS patients can communicate using the P300 BCI; however, no study has systematically examined BCI performance longitudinally. This study examines the relationship between disease progression, BCI performance, and several other factors. Nine patients are enrolled in the study and data are collected in intervals of eight weeks. In each session, the ALSFRS-r is administered to track disease progression, followed by four EEG tasks: resting EEG, a cognitive task, an oddball task, and a copy spelling task. The oddball task generated a P300 for attended stimuli, and not for unattended stimuli. ALSFRS-r scores have decreased by 2.8 points on average, mean BCI copy spelling accuracy is 94.4%. Overall, the results suggest that BCI performance remains stable as disease progression continues.