John E. Richards Home Page
Abstract

Guy. M., Richards, J.E. & Roberts, J.E. (2018). Accurate head models for cortical source analysis in infants at high risk of autism spectrum disorders. International Conference on Infant Studies, Philadelphia, July, 2018.(pdf )

The aim of this study was to develop the first realistic head models for use with infants at high risk of autism spectrum disorders (ASD) and to test these head models in the source analysis of electrophysiological data collected during face processing. The optimal approach for source analysis involves using realistic head models based upon individual participants’ structural MRIs, however, this is not always feasible. Careful selection of alternative head models is important to accurate source localization, and may be critical when examining infants at high-risk of neurodevelopmental disorders. Twelve-month-old participants included two high-risk groups, 21 infant siblings of children with ASD (ASIBs) and 15 infants diagnosed with fragile X syndrome (FXS), and 21 typically developing, low-risk control (LRC) infants. All participants completed a face processing ERP experiment. Structural MRIs were collected from a subset of the participants. Realistic head models were created from the MRIs; materials within the head were identified, segmented, and assigned a relative conductivity. Current density reconstruction (CDR) of the N290 ERP component was done with head models created from participants’ own MRIs to examine activation in regions of interest (ROIs) believed to be highly relevant to face processing. CDR activity was analyzed in an ANOVA including participant group, stimulus type, and ROI. There were main effects of stimulus type, F(1, 38) = 17.01, p = 0.0002, reflecting greater activation to faces than toys, and of ROI, F(17, 646) = 20.43, p < 0.0001, reflecting high levels of activation in the middle fusiform gyrus and anterior temporal brains areas. Subsequent source analyses were completed with ASIB and FXS groups to determine the impact of head model on CDR activation. We tested head models created from infants’ own MRIs against those created from the average of study- and group-specific MRIs, group-specific MRIs obtained from the Infant Brain Imaging Study (IBIS; 25 FXS MRIs, 53 ASIB MRIs), and MRIs collected from TD infants. An ANOVA tested effects of head model, participant group, and ROI on CDR activity. There was a significant effect of head model, F(3, 51) = 13.30, p < 0.0001, reflecting similar CDR across own-MRI and IBIS head models, but not study-specific and TD head models. There was an interaction of head model and stimulus type, F(3,51) = 3.88, p = 0.0141, reflecting similar differentiation of responses to faces and toys in the own-MRI and IBIS head models, and less differentiation in study-specific and TD head models. An interaction of head model and ROI, F(51, 867) = 6.23, p < 0.0001, indicated that while the IBIS head model was the best alternative to infants’ own MRIs, the quality of fit varied across ROI. Results indicate that head model selection is important to accurate source analysis and may be complex in high risk groups. IBIS head models proved the best match to infants’ own MRIs, possibly due to greater heterogeneity in the brains of infants at high risk of ASD that was better accounted for in a model created from a large collection of group-specific MRIs.