Investigation of cooperative cortical dynamics by multivariate autoregressive modeling of event-related local field potentials

Abstract

To explore the operation of large-scale networks in the cerebral cortex, we sought to measure the functional interdependence of event-related local field potentials (LFPs) from different cortical areas in macaque monkeys performing a visuomotor pattern discrimination task. To track the transformation of functional interdependence accompanying rapid changes in cognitive state, we developed a method for spectral coherence analysis using multivariate autoregressive (MVAR) models of short-windowed LFP time series. MVAR modeling overcomes problems associated with direct coherence analysis of short-windowed data. Coherence and phase are shown to vary during task processing with spatial location, processing stage, and stimulus and response conditions.