Poster Session D, Monday, March 25, 8:00 – 10:00 am, Pacific Concourse
Investigating individual variation in cognitive function through Mesoscale Individualized Neurodynamic (MINDy) models
Matthew F. Singh1, ShiNung Ching1, Todd S. Braver1; 1Washington University in St. Louis
It is widely thought that a critical aspect of individual variation in human cognition relates to how information is transferred between widely-distributed regions, in addition to within-region local computations. Although it is possible to robustly characterize individual differences in brain network functional connectivity (FC) in a data-driven manner, such approaches don’t allow for the overt construction of mechanistic hypotheses. By contrast, generative biologically-plausible models do enable testing of mechanistic hypotheses, but to date these have not been individually constructed directly from functional data. In the current work, we bridge this gap with a novel method to fit high-resolution Mesoscale Individualized Neurodynamic (MINDy) models directly to fMRI data. The estimation procedure enables networks consisting of hundreds to thousands of nonlinear neural masses (parcels) to be fit individually for each subject, requiring only a minimum of 15 minutes worth of resting-state data. Model parameters are highly reliable across scanning sessions and robust to preprocessing choices. Ground-truth tests also indicate that recovered models are valid and robust to both additive measurement noise and hemodynamic uncertainty. The model mechanistically reveals sources of individual variation in brain networks that may be mis-estimated in standard resting-state FC approaches. The findings represent an advance on two fronts: generating spatially-detailed biophysically plausible models at the level of individual human brains which utilize a new measure of connectivity. Planned follow-up analyses will apply the model to examine how cognitive task states modify brain network dynamics and provide new metrics for characterizing cognitive individual differences.
Topic Area: METHODS: Neuroimaging