Characterizing the Heterogeneity in Autism Spectrum Disorder Using Brain Connectivity Underlying Social Cognition

Friday, May 12, 2017: 5:00 PM-6:30 PM
Golden Gate Ballroom (Marriott Marquis Hotel)
M. Thye1 and R. K. Kana2, (1)Psychology, University of Alabama at Birmingham, Birmingham, AL, (2)University of Alabama at Birmingham, Birmingham, AL
Background: Autism Spectrum Disorder (ASD) is characterized by deficits in theory-of-mind (ToM) which negatively impact social interaction. Previous neuroimaging studies have found alterations in ToM brain network in individuals with ASD. However, most of these studies rely on standard fMRI analyses which concatenate results to arrive at a group-level model that may not be reflective of many or most of the participants within a heterogeneous group. A novel approach to studying functional connectivity that accounts for the heterogeneity of ToM regions in an ASD population is the Group Iterative Multiple Model Estimation (GIMME) algorithm which reveals divergent subgroups based on patterns of functional connectivity among a priori regions of interest (ROI) using a structural equation modeling framework.

Objectives: To examine the nature and extent of neural heterogeneity across ASD and typically developing (TD) participants in an fMRI study of ToM processing.

Methods: A total of 63 participants (32 ASD and 31 age-and-IQ-matched TD) between the ages of 10 and 36 years took part in this fMRI study. In the scanner, participants watched animations of geometrical shapes depicting ToM. ROIs were derived from a Neurosynth mask based on previous studies of ToM and included: left and right inferior frontal gyrus (LIFG; RIFG), left precuneus (LPCUN), left and right posterior superior temporal sulcus (LpSTS; RpSTS), and medial prefrontal cortex (MPFC). Based on the fMRI time-series information extracted from each ROI during the ToM condition, GIMME identifies the presence and direction of connections among the ROIs to arrive at a group level model that is then tested at the individual level to determine if subgroups exist within the data that differ from the group model.

Results: A group connectivity map was found with connections from MPFC to LIFG, LpSTS to MPFC and to LPCUN, and RpSTS to LpSTS. Within the ToM condition, two functional connectivity-based subgroups were identified: Subgroup A comprising 28% of the ASD and 39% of the TD participants was characterized by increased connectivity from LPCUN to RpSTS as well as increased connectivity from MPFC to RIFG. Conversely, Subgroup B which contained 72% of the ASD and 61% of the TD participants showed comparatively weaker connectivity with no additional pathways emerging above the group level model. Statistical comparisons of the individuals comprising the two subgroups revealed stronger connectivity of the group level connection from LpSTS to MPFC in Subgroup A compared to Subgroup B (t61 = -3.118, p< .05). No group level paths were stronger for Subgroup B.

Conclusions: The pattern of results within the ToM condition suggests possible underconnectivity in the group containing the largest percentage of ASD participants. The GIMME algorithm, which is blind to diagnostic classification, was able to detect the presence and direction of connections within a heterogeneous sample during a task-based fMRI study of ToM. The findings of this study are preliminary, and there are several follow-up analyses planned. These results should be examined further in the context of the diagnostic and demographic makeup of the subgroups.