Abstract
Introduction:
The breadth of information afforded by multimodal brain imaging can substantially advance our understanding of brain network function. However, the inherent properties of each imaging modality have imposed conceptual and technical challenges on development of integrative approaches. Measures of white-matter connectivity derived from diffusion MRI have been previously shown to usefully inform resting-state functional connectivity (Honey et al. 2009) and task-related effective connectivity in a four-node network (Stephan et al. 2009). Here, we developed an integrative analysis of structural and effective connectivity applicable to larger scale networks. Furthermore, we intended to evaluate how modelled indirect structural connectivity relates to mainly polysynaptic effective connectivity.
Methods:
Functional MRI and high angular resolution diffusion imaging (HARDI) were performed with a 3T MRI scanner (TimTrio, Siemens Medical Solutions, Erlangen, Germany) in 17 healthy adult participants who had to recognize point-light body motion displays. Pre-processing and analysis were conducted with Statistical Parametric Mapping (SPM12, http://www.fil.ion.ucl.ac.uk/spm) for fMRI and the FMRIB Software Library (FSL5, http://www.fmrib.ox.ac.uk/fsl) for HARDI data. The structural connection strengths provided by probabilistic tractography on the HARDI data were averaged across individuals and introduced as prior probabilities for corresponding second-level effective connections in dynamic causal modelling (DCM). Furthermore, to model indirect structural connectivity, graph theoretical approaches were applied to the group structural adjacency matrix.
Results:
Bayesian model reduction (Friston et al. 2016) provided very strong evidence in favour of effective connectivity models informed by structural connectivity as compared to uninformed ones (Sokolov et al. in press). Modelling high-order, indirect structural connectivity resulted in additional substantial improvement in the evidence of effective connectivity models, suggesting that ensemble dynamics are determined by the sensitivity of network nodes to their afferents.
Conclusions:
The findings indicate that integration of anatomical information afforded by diffusion MRI substantially improves models of effective connectivity. Modelling of asymmetric high-order structural connectivity more usefully informs effective connectivity than conventional symmetric measures of direct structural connectivity. In conjunction with connectivity-based prediction of behaviour, these integrative approaches may provide novel insights into the structure-function relationships and dynamics of brain networks, both in normalcy and in neuropsychiatric conditions.
