
Synthetic PET: Analyzing large‐scale properties of neural networks
Author(s) -
Arbib M. A.,
Bischoff A.,
Fagg A. H.,
Grafton S. T.
Publication year - 1994
Publication title -
human brain mapping
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.005
H-Index - 191
eISSN - 1097-0193
pISSN - 1065-9471
DOI - 10.1002/hbm.460020404
Subject(s) - neurophysiology , artificial intelligence , artificial neural network , neural activity , computer science , subnetwork , neuroscience , computational model , saccade , human brain , synthetic data , machine learning , pattern recognition (psychology) , psychology , eye movement , computer security
Synthetic PET is introduced as a new computational technique for connecting neural network studies based on animal data and image studies of human brain function. Synthetic PET comparisons are taken from a computational model of interacting neural networks in, for example, the monkey brain by integrating synaptic activity in each subnetwork as different simulated tasks are performed. Given a pair of tasks, comparative synaptic activity levels for each modeled neural region are then painted into the homologous regions of a three‐dimensional model of the human brain corresponding to the Talairach atlas. The resulting comparison then offers predictions of relative changes of neuronal activity as obtained from PET comparisons of humans performing a similar pair of tasks. Where neurophysiological data and individual arrays of neural networks give us a very fine‐grained view of the behavior of just a few neurons within nonhumans, PET measures provide a more global view of brain function. Synthetic PET allows us to integrate human and monkey studies in the construction of more complete models of the neural mechanisms used by both species in similar tasks. The technique is demonstrated in a saccade generation task. The method readily generalizes to other forms of brain imaging. © 1995 Wiley‐Liss, Inc.