Download 46340-2-12118

International Biometric Society
Time-dependent network structure in fMRI data
Lucy F. Robinson, Lauren Y. Atlas and Tor D. Wager
Drexel University, National Institutes of Health, University of Colorado, Boulder
We present a new method for detecting time-dependent structure in networks of
brain regions using a hidden Markov stochastic blockmodel. Most analyses of
functional connectivity in fMRI data assume that network behavior is static in time, or
differs between task conditions with known timing. However, in resting state data or
experiments with drug uptake, learning, or complex cognitive tasks, shifts in network
behavior may occur at unknown times. Our proposed method identifies distinct
functional connectivity states with respect to the organization of the functional
connectivity network into communities of highly connected brain regions, as
characterized by a stochastic blockmodel. A change in the structural organization of
the functional connectivity network could consist of, for example, a shift from a state
in which connectivity is highly modular, i.e. connectivity is concentrated within
subnetworks of regions, to a state in which the pattern of connectivity is less modular
and more integrated throughout the network. Changes in network structure may be
related to shifts in neurological state. We apply this approach to data from an
experiment examining how contextual factors influence drug-induced analgesia.
International Biometric Conference, Florence, ITALY, 6 – 11 July 2014