Underlying Features of Inter-Individual Variability in Brain Function Using a Siamese Neural Network

Keywords
Loading...
Thumbnail Image
Issue Date
2021-08-31
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
Large-scale networks are comprised of distributed brain areas that show correlated fluctuations in their spontaneous resting state functional magnetic resonance imaging activity, and these patterns of functional connectivity show high individual variability. However, it is still unknown what the most important dimensions are along which individual functional connectivity patterns can vary. In this study, we investigated the discriminating inter-individual features of brain function. We used a deep neural network to learn to discriminate subjects’ connectivity patterns from each other (n = 996). We found that we can reduce the high-dimensional connectivity profiles to only three components and still identify subjects’ functional connectivity patterns with 96% accuracy. The three identified components were mapped back onto the large-scale brain networks, and were classified as 1) sensorimotor networks component, 2) within-networks component, 3) higher-order networks component. Furthermore, we found correlations between the higher-order networks component and cognitive measures as well as with relevant brain regions’ area size. This implies that we identified three robust components of functional connectivity that can be linked to cognition and amount of grey matter.
Description
Citation
Faculty
Faculteit der Sociale Wetenschappen