Network Sculpting Index Suggests Impaired Functional Network Differentiation in ASD

Background: There is increasing consensus that ASD is a disorder involving distributed brain networks. Functional connectivity MRI (fcMRI) studies have produced mixed findings, including both underconnectivity and overconnectivity in ASD, implicating default mode network (DMN), mirror neuron system (MNS), and the mentalizing (Theory of Mind, ToM) network. Seemingly contradictory findings could reflect complementary aspects of impaired network sculpting, characterized by decreased connectivity within neurotypical networks and increased connectivity outside of these networks. 

Objectives: To evaluate the efficiency of neurotypical networks (DMN, MNS, ToM) in ASD and typically developing (TD) groups by calculating a network sculpting index (NSI), a ratio of within network connectivity and outside network connectivity.

Methods: Resting-state functional MRI data were acquired for 6:10 minutes for 33 children and adolescents with ASD (7-18 y/o) and 33 TD participants matched for age, motion, and non-verbal IQ. Standard preprocessing involved motion and field map correction, spatial smoothing, low bandpass filtering (.008<f<.08Hz), nuisance regression, and standardization to MNI152 template. For each network (DMN, MNS, ToM), regions of interest (ROIs) were identified using 6mm-radius spheres centered on previously reported coordinates (Van Overwalle & Baetens, 2009; Watanabe et al., 2012). Within-network masks were created for each ROI, including the seed and all other ROIs in the network (dilated to 12mm radius); outside-network masks excluded all network ROIs and non-cortical voxels. Individual whole-brain correlation maps (using average time series extracted from seed) were cluster corrected (p<1^-6) and Fisher-transformed to z’. For each network, the number of significant voxels, weighted by z’, was determined for within- and outside-network masks, and was then used to calculate the NSI using the formula: NSI = (WNC-ONC)/(WNC+ONC), with WNC and ONC being within and outside-of network connectivity, respectively.

Results: Mean NSI was significantly reduced in the ASD group (p<.05) for DMN whereas no significant differences were detected for MNS or ToM. A correlational analysis of NSI and three a priori selected social measures (ADI-Social, ADI-Communication, and ADOS Communication + Social [CS] scores available for ASD participants only) revealed a negative relationship between NSI and ADOS CS in MNS (r=-0.35, p<.05) and ADI-Communication scores in ToM (r=-.39, p<.05). Based on these correlations, a post hoc analysis was performed in a subset of ASD participants (n=25) with highest level of social symptomatology as defined by ADOS CS scores ≥10. Direct group comparison of this ASD subsample and 25 TD participants matched on age, motion, and non-verbal IQ corroborated earlier results in DMN (TD>ASD, p<.05). In addition, this analysis yielded significant between-group differences in mean NSI for MNS and ToM (TD>ASD, p<.05). 

Conclusions: Findings suggest that DMN is less efficiently sculpted in individuals with ASD, compared to TD controls (more connectivity outside of the network than within). Concordant between-group differences were also found for MNS and ToM networks, but these were detected only in ASD participants with relatively high symptom severity.