Speaker
Description
Estimating whole-brain directed connectivity from resting-state fMRI remains an open challenge. Dynamic Causal Modeling (DCM) is prohibitive at whole-brain scale, while lagged correlations conflate direct and indirect propagation with common-input artifacts. We recover directed connectivity by fitting a Network Diffusion Model (NDM) to temporal dynamics. The NDM \cite{Abdelnour2014} describes propagation on connectome $S$ via $d\mathbf{x}/dt=-\beta(I-S)\mathbf{x}(t)$. Replacing symmetric $S$ with unknown directed $E$, model-predicted lagged covariances:
$$\mathrm{FC}(\tau;E)=e^{-\beta t(I-E)}\left(e^{-\beta(t+\tau)(I-E)}\right)^T$$ are asymmetric for $\tau>0$, encoding directional flow. We estimate $E$ by fitting empirical lagged covariances across multiple lags, with sigmoid regularization constraining edges to structural pathways while permitting asymmetry. Applied to resting-state fMRI from 300 HCP \cite{VanEssen2013} subjects (100 cortical \cite{Schaefer2018} + 14 subcortical regions), the directed network was sparser than empirical lagged FC, with all edges supported by structural connections. Net flow revealed consistent hierarchical organization across subjects: subcortical structures, V1, and limbic areas as sources, and somatomotor and dorsal attention areas as sinks. These findings show that temporal precedence in lagged correlations can be misleading, and that modeling propagation through network topology recovers directed pathways grounded in anatomy.
Bibliography
@article{Abdelnour2014,
author = {Abdelnour, Farras and Voss, Henning U. and Raj, Ashish},
title = {Network diffusion accurately models the relationship between structural and functional brain connectivity networks},
journal = {NeuroImage},
volume = {90},
pages = {335--347},
year = {2014},
doi = {10.1016/j.neuroimage.2013.12.039}
}
@article{VanEssen2013,
author = {Van Essen, David C. and Smith, Stephen M. and Barch, Deanna M. and Behrens, Timothy E.J. and Yacoub, Essa and Ugurbil, Kamil},
title = {The {WU-Minn} {Human Connectome Project}: An overview},
journal = {NeuroImage},
volume = {80},
pages = {62--79},
year = {2013},
doi = {10.1016/j.neuroimage.2013.05.041}
}
@article{Schaefer2018,
author = {Schaefer, Alexander and Kong, Ru and Gordon, Evan M. and Laumann, Timothy O. and Zuo, Xi-Nian and Holmes, Avram J. and Eickhoff, Simon B. and Yeo, B.T. Thomas},
title = {Local-global parcellation of the human cerebral cortex from intrinsic functional connectivity {MRI}},
journal = {Cerebral Cortex},
volume = {28},
number = {9},
pages = {3095--3114},
year = {2018},
doi = {10.1093/cercor/bhx179}
}
