Large-scale neural model validation of partial correlation analysis for effective connectivity investigation in functional MRI

Inserm, u678, CHU Pitié-Salpêtrière, Paris F-75013, France.
Human Brain Mapping (Impact Factor: 5.97). 03/2009; 30(3):941-50. DOI: 10.1002/hbm.20555
Source: PubMed


Recent studies of functional connectivity based upon blood oxygen level dependent functional magnetic resonance imaging have shown that this technique allows one to investigate large-scale functional brain networks. In a previous study, we advocated that data-driven measures of effective connectivity should be developed to bridge the gap between functional and effective connectivity. To attain this goal, we proposed a novel approach based on the partial correlation matrix. In this study, we further validate the use of partial correlation analysis by employing a large-scale, neurobiologically realistic neural network model to generate simulated data that we analyze with both structural equation modeling (SEM) and the partial correlation approach. Unlike real experimental data, where the interregional anatomical links are not necessarily known, the links between the nodes of the network model are fully specified, and thus provide a standard against which to judge the results of SEM and partial correlation analyses. Our results show that partial correlation analysis from the data alone exhibits patterns of effective connectivity that are similar to those found using SEM, and both are in agreement with respect to the underlying neuroarchitecture. Our findings thus provide a strong validation for the partial correlation method.

13 Reads
  • Source
    • "In this case, the correlations between neural units is inferred once the effects of all other units have been removed. Partial correlations are typically preferred to Pearson's correlation coefficient as they have been shown to be better suited to detecting changes in connectivity structure [Smith et al., 2011, Marrelec et al., 2009]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Understanding the functional architecture of the brain in terms of networks is becoming increasingly common. In most fMRI applications functional networks are assumed to be stationary, resulting in a single network estimated for the entire time course. However recent results suggest that the connectivity between brain regions is highly non-stationary even at rest. As a result, there is a need for new brain imaging methodologies that comprehensively account for the dynamic (i.e., non-stationary) nature of the fMRI data. In this work we propose the Smooth Incremental Graphical Lasso Estimation (SINGLE) algorithm which estimates dynamic brain networks from fMRI data. We apply the SINGLE algorithm to functional MRI data from 24 healthy patients performing a choice-response task to demonstrate the dynamic changes in network structure that accompany a simple but attentionally demanding cognitive task. Using graph theoretic measures we show that the Right Inferior Frontal Gyrus, frequently reported as playing an important role in cognitive control, dynamically changes with the task. Our results suggest that the Right Inferior Frontal Gyrus plays a fundamental role in the attention and executive function during cognitively demanding tasks and may play a key role in regulating the balance between other brain regions.
    NeuroImage 10/2013; 103. DOI:10.1016/j.neuroimage.2014.07.033 · 6.36 Impact Factor
  • Source
    • "probabilistic and global fiber-tracking was performed to analyze structural connectivity. In a comparison study of different network modeling techniques (Smith et al., 2011) and in direct comparison with structural equation modeling (Marrelec et al., 2009), partial correlations performed well in reconstructing networks based on time series data. Partial correlations are commonly used to analyze direct relationships between time signals, whereas full correlations are more prone to introduce indirect connections into the model. "
    [Show abstract] [Hide abstract]
    ABSTRACT: An emerging field of human brain imaging deals with the characterization of the connectome, a comprehensive global description of structural and functional connectivity within the human brain. However, the question of how functional and structural connectivity are related has not been fully answered yet. Here, we used different methods to estimate the connectivity between each voxel of the cerebral cortex based on functional magnetic resonance imaging (fMRI) and diffusion tensor imaging (DTI) data in order to obtain observer-independent functional-structural connectomes of the human brain. Probabilistic fiber-tracking and a novel global fiber-tracking technique were used to measure structural connectivity whereas for functional connectivity, full and partial correlations between each voxel pair's fMRI-timecourses were calculated. For every voxel, two vectors consisting of functional and structural connectivity estimates to all other voxel's in the cortex were correlated with each other. In this way, voxels structurally and functionally connected to similar regions within the rest of the brain could be identified. Areas forming parts of the 'default mode network' (DMN) showed the highest agreement of structure-function connectivity. Bilateral precuneal and inferior parietal regions were found using all applied techniques, whereas the global tracking algorithm additionally revealed bilateral medial prefrontal cortices and early visual areas. There were no significant differences between the results obtained from full and partial correlations. Our data suggests that the DMN is the functional brain network, which uses the most direct structural connections. Thus, the anatomical profile of the brain seems to shape its functional repertoire and the computation of the whole-brain functional-structural connectome appears to be a valuable method to characterize global brain connectivity within and between populations.
    NeuroImage 10/2013; DOI:10.1016/j.neuroimage.2013.09.069 · 6.36 Impact Factor
  • Source
    • "In this context, deficits in functional integration or connectivity are implied when the influence of one brain region on another is stronger or weaker in patients relative to control subjects (Price et al., 2006; Ween, 2008). There are several approaches that have been proposed to assess functional integration or connectivity (for a review see Ramnani et al., 2004; Harrison et al., 2007; Rogers et al., 2007; Li et al., 2009; Bressler and Menon, 2010), including structural equation modeling (McIntosh and Gonzalez-Lima, 1994), dynamic causal modeling (DCM) (Friston et al., 2003), Granger causality (Roebroeck et al., 2005), psycho-physiological interactions (Friston et al., 1997), dynamic Bayesian networks (Rajapakse and Zhou, 2007), multivariate autoregressive modeling (Harrison et al., 2003), partial correlation analysis (Marrelec et al., 2009), non-linear system identification (Li et al., 2010b), and switching linear dynamic systems (Smith et al., 2010). Each method has its advantages and weaknesses (e.g., see Penny et al., 2004; Ramnani et al., 2004; Witt and Meyerand, 2009) and its use should be motivated by the question of interest, level of inference, paradigm design, data acquisition and analysis. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Functional imaging studies of brain damaged patients offer a unique opportunity to understand how sensorimotor and cognitive tasks can be carried out when parts of the neural system that support normal performance are no longer available. In addition to knowing which regions a patient activates, we also need to know how these regions interact with one another, and how these inter-regional interactions deviate from normal. Dynamic causal modeling (DCM) offers the opportunity to assess task-dependent interactions within a set of regions. Here we review its use in patients when the question of interest concerns the characterization of abnormal connectivity for a given pathology. We describe the currently available implementations of DCM for fMRI responses, varying from the deterministic bilinear models with one-state equation to the stochastic non-linear models with two-state equations. We also highlight the importance of the new Bayesian model selection and averaging tools that allow different plausible models to be compared at the single subject and group level. These procedures allow inferences to be made at different levels of model selection, from features (model families) to connectivity parameters. Following a critical review of previous DCM studies that investigated abnormal connectivity we propose a systematic procedure that will ensure more flexibility and efficiency when using DCM in patients. Finally, some practical and methodological issues crucial for interpreting or generalizing DCM findings in patients are discussed.
    Frontiers in Systems Neuroscience 08/2010; 4. DOI:10.3389/fnsys.2010.00142
Show more