Abstracts

Rationale: In development of brain-computer interface methods, a previous study (Shih et al, 2009) determined that the P300 response to novel stimuli underlies letter discrimination ability in subjects during electrocortigraphy (ECoG) recordings. Localization of the response, however, does not appear to respect traditional neuroanatomic localization, with responses recorded from diverse dominant hemisphere locations. We hypothesize that neuronal networks rather than specific foci underlay visual character recognition. Methods: Three subjects with intractable epilepsy underwent intracranial subdural and depth electrode recordings. Alphanumeric characters via a 6x6 square matrix were presented to subjects who were instructed to silently count the number of times a target character was highlighted. >100 samples of ECoG recorded from subsets of electrodes were recorded during true character recognition ( ), false character recognition ( - ), and base states without character highlighting. Network activity was measured iteratively for all electrode pairs with the synchrony index (SI), a measure of local amplitude coherence combined with electrode pair phase synchrony linkage. Electrode pairs with SI values falling above or below the 95th confidence limit (determined via t-scores) of the mean difference between ( ) and (-) states compared to the base state were mapped and compared to regions previously established as important in the P300 response. Results: In all three subjects, during ( ) states a nexus of relative desynchronization extended beside regions of P300 responses. The same general region of desynchronization, usually more limited in extent, was maintained during (-) states, with emergence of focal regions of hypersynchrony. Regions of hyper- or desynchronization usually did not involve electrodes with consistent P300 responses. In one case with the entire grid available for mapping, the difference between ( ) and (-) states was significantly different demonstrating regional hypersynchrony during (-) states. The two other cases that lacked full electrode assessment did not demonstrate significant differences between recognition states. Conclusions: Regional networks of synchronous neuronal activity may accompany the P300 responses that underlie simple character recognition tasks. Neuronal network properties and the means to quantify their activity may be important in the design of brain-computer interfaces.