Granger causality reveals abnormal network structure in the right hemisphere in children with high-functioning autism

Background: Recent neuroimaging studies have suggested that children with autism spectrum disorder (ASD) have abnormal functional connectivity as compared to typically developing (TD) controls, but few have investigated whether the hierarchical structure is different between these two groups. Recently we have used granger causality to demonstrate differences in hierarchical structure within the phonological language network during the period just before presentation of a visual phonological stimulus in dyslexic v typical readers and demonstrated how variations within the hierarchical structure are associated with higher function in the dyslexic group. 

Objectives: To investigate differences in functional network structure between ASD and TD children using our established paradigm. 

Methods: Ten children with high-functioning ASD and 10 TD controls matched on age and gender performed several visual and auditory phonological tasks during a magnetoencephalography scan. Structural MRI scans and neuropsychological assessments were also obtained for each subject. Sources were localized on the cortical surface using the minimum norm estimate (MNE) method. Data from 500ms prior to the onset of the stimuli were extracted from six cortical regions [right and left inferior frontal area (IFA), temporoparietal area (TPA) and visual word form area (VWFA)] and filtered into gamma, beta, alpha, delta and theta frequency bands. The five sources from each cortical area with the highest activity were selected for analysis. Granger causality was calculated using Dynamic Autoregressive Neuromagnetic Causal Imaging (DANCI). An analysis of variance conducted on the connectivity values for each region was used to determine significant differences. Alpha was corrected to 0.001 to account for multiple analyses. Data from six ASD and five TD participants were available for this analysis. 

Results: Differences in connectivity were seen in both the left and right hemisphere but changes in the hierarchical network structure were only found in the right hemisphere.

Analysis for the left VWFA demonstrated an ASD by brain area interaction for the gamma frequency range. This resulted from greater connectivity between the left VWFA and TPA for the ASD participants as compared to the TD participants but not other regions. Analysis of the right TPA demonstrated an ASD by brain area by connectivity direction interaction for the beta range and delta frequency band. This was driven by significant differences in connectivity direction between the right TPA and IFA for the ASD and TD groups. In addition, Analysis of the right IFA demonstrated an ASD by brain area by connectivity direction interaction for the alpha frequency band. This was also driven by significant differences in connectivity direction between the right TPA and IFA for the ASD and TD groups.

Conclusions: Although preliminary, the results implicate that the hierarchical structure of the right hemisphere is different between ASD and TD participants, particularly with regards to the functional organization of the pathways between the right IFA and right TPA. This may point to deficient top-down control of the TPA from the IFA. Analysis of further participants and additional paradigms will help clarify these differences as will inclusion of performance measures as cofactors in our analysis.