Time-Varying Temporal Functional Modes: An instantaneous connectivity-based method that probes brain network reconfigurations while allowing for a shared anatomical infrastructure

Thumbnail Image
Issue Date
Journal Title
Journal ISSN
Volume Title
The core cognitive neuroscientific aim is to determine how the brain gives rise to mental function. The answers provided to this ontological question depend, in part, on predominant methodology. Initial functional magnetic resonance imaging studies were mainly segregationist, i.e., focused on mapping psychological functions to individual brain sites. The modularity hypothesis underlying this approach, however, has been questioned since its conception. Conversely, it is argued that probing the neural underpinnings of mental function requires examination of interactions within and between integrated, spatially distributed functional brain networks. This is for example accomplished using functional connectivity methods, which examine co-variation of spatially distributed signals emitted from the brain. Time-averaged estimations of functional connectivity are commonly obtained, which assumes that the dependence structure between brain regions is constant over the course of an fMRI scan. In the past decennia, a line of research aims to probe time-varying functional connectivity changes of these networks, based on the premise that biologically meaningful changes in inter-regional relations are expected to occur. Available time-varying functional connectivity methods do not allow for spatial overlap, i.e., for functional networks to share a common anatomical infrastructure. This is addressed in Smith et al. (2012), where a spatial independent component analysis and a temporal decomposition are applied consecutively, resulting in so-called "Temporal Functional Modes" (TFMs). Classical TFM analysis is time-invariant, however, in the sense that the extent to which TFMs recruit specific spatial components (e.g. canonical functional connectivity networks) is assumed to be constant. To capture the time-varying nature of functional connectivity and simultaneously allow for spatial overlap, we developed a novel, time-varying version of classical TFM (TV-TFM) analysis. With this extension, we obtain moment-by-moment estimations of brain network reconfigurations, i.e. of how statistically independent temporal processes (i.e. TFMs) recruit statistically independent spatial components. Properties of the new model were explored by means of data simulations. Proof of principle for this new method was obtained by application to high temporal resolution fMRI data of participants performing a visuomotor association task. It was shown that the brain network reconfigurations captured with the new method closely followed the expectation based on the task design. Keywords: time-varying functional connectivity, dynamic functional connectivity, independent component analysis, temporal functional modes, spatial overlap
Faculteit der Sociale Wetenschappen