Instantaneous interactions between brain sites can distinguish movement from rest but are relatively poor at resolving different movement types

Classification of finger movements from brain signals is an important problem in emerging neural prosthetics research. Current techniques have focused primarily on aggregating data from different brain sites independently, without regard to the interaction between different brain sites. We performed a very simple experiment classifying finger movements during electrocorticographic recording from motor cortex. Unlike previous experiments of this type, we examined whether the most-simple type of instantaneous interactions between brain regions could improve classification. We focused on two simple, but salient markers: phase coherence of the 12-20 Hz beta range, and simple correlation in broadband spectral change. We compared behavioral classification using these interactions with simple amplitude changes in beta and broadband magnitude. We found that inter-electrode phase coherence and cross-correlation, were relatively poor at distinguishing between finger movements, and only mediocre at identifying null/rest periods.