a correlation method was developed to examine functional interactions between brain regions by correlating regional cerebral blood flows (rCBF) or cerebral metabolic rates for glucose (rCMRglc) determined by positron emission tomography (PET) in humans. When rCBF was measured during a face matching task in which task difficulty was systematically changed (by adding noise to the images), brain functional interactions changed from regions doing predominantly perceptual processing to frontal lobe regions. In a study of response time during a motor preparation task, cerebellar rCBF was negatively correlated with response time. A systems-level neural network model, fitted to rCBF PET data, permitted determination of brain regions that were involved in a working memory for faces task. Such a network obtained from PET data in young subjects performing a face-matching task provided a good fit to data for old subjects but not for mildly demented Alzheimer's disease (AD) patients, although patients performed the task with the same accuracy as controls. A large-scale neural model was developed to relate PET data to temporal and spatial activity of neuronal populations during specific cognitive tasks. Simulations of selective attention for features were performed.
