Dynamic Bayesian network modeling of fMRI: A comparison of group-analysis methods

Abstract

Bayesian network (BN) modeling has recently been introduced as a tool for determining the dependencies between brain regions from functional-magnetic-resonance-imaging (fMRI) data. However, studies to date have yet to explore the optimum way for meaningfully combining individually determined BN models to make group inferences. We contrasted the results from three broad approaches: the "virtual-typical-subject" (VTS) approach which pools or averages group data as if they are sampled from a single, hypothetical virtual typical subject; the "individual-structure" (IS) approach that learns a separate BN for each subject, and then finds commonality across the individual structures, and the "common-structure" (CS) approach that imposes the same network structure on the BN of every subject, but allows the parameters to differ across subjects. To explore the effects of these three approaches, we applied them to an fMRI study exploring the motor effect of L-dopa medication on ten subjects with Parkinson's disease (PD), as the profound clinical effects of this medication suggest that fMRI activation in PD subjects after medication should start approaching that of age-matched controls. We found that none of these approaches is generally superior over the others, according to Bayesian-information-criterion (BIC) scores, and that they led to considerably different group-level results. The IS approach was more sensitive to the normalization effect of the L-dopa medication on brain connectivity. However, for the more homogeneous control population, the VTS approach was superior. Group-analysis approaches should be selected carefully with consideration of both statistical and biomedical evidence.

Graphical Models of Functional MRI Data for Assessing Brain Connectivity

“…Proposed methods include correlation thresholding (Cao & Worsley, 1999), linear decomposition (Calhoun et al, 2001;McKeown, 2000), structural equation models (SEM) (Bollen, 1989), multi-variate auto-regression (Valdes-Sosa et al, 2005), dynamic causal models (Friston et al, 2003), Bayesian networks (Li et al, 2008;Zheng & Rajapakse, 2006), wavelet analysis (Bullmore et al, 2004), and clustering (Heller et al, 2006). Correlation thresholding (Cao & Worsley, 1999) directly examines the correlation between the activities of brain regions.…”

“…A review of the literature shows that current group-analysis methods based on graphical models can be classified into three broad categories (see Fig. 7), as discussed as follows (Li et al, 2008). First, we could ignore subject diversity, and assume that the brains of all the subjects are structured and function in a similar way, as if there is a virtual typical subject able to satisfactorily represent the whole group.…”

“…The most straight-forward implementation of the "individual-model approach" is to directly input separately learned individual models as subject features into a second-level analysis. For example, structural features of the network of individual models can be selected with classification and cross-validation procedures at the group level, as applied in (Li et al, 2008). A theoretically elegant approach is to build a group-level model to describe the diversity distribution in a group, and the group-level and the individual-level models form a big model over the group data.…”

A Spectral Graphical Model Approach for Learning Brain Connectivity Network of Children's Narrative Comprehension

“…Instead, the optimization of the network in the frequency domain can generate new graphs that include connections that were not predicted a priori. Moreover, compared with other BN (BN, DBN) that have been applied to the structure learning of neural networks using fMRI data, the SBN approach considers the whole spectrum (and thus the whole autocovariance function, G(h), h ‡ 0) of the fMRI time series in the approximation of the likelihood function, while the applications using other BN only considered fixed lags such as G(1) and G(0) (Li et al, 2008;Rajapakse and Zhou, 2007;Zheng and Rajapakse, 2006).…”

“…The Dynamic Bayesian Network (DBN) is an extension of BN and provides a framework for building networks on time-series data using fixed-length time-delayed edges in the graph. In Burge et al (2009), Li et al (2008), and Rajapakse and Zhou (2007), DBN was applied to analyzing fMRI data for various cognitive tasks.…”