Abnormal Functional Connectivity in the Social Brain Identified By a Generalized Classifier for Autism Spectrum Disorder

Thursday, May 11, 2017: 5:30 PM-7:00 PM
Golden Gate Ballroom (Marriott Marquis Hotel)
R. Hashimoto, Medical Institute of Developmental Disabilities Research, Showa University, Setagaya-ku, Japan; Department of Language Sciences, Graduate School of Humanities, Tokyo Metropolitan University, Tokyo, Japan
Background: Autism spectrum disorder (ASD) has been increasingly conceptualized as a disease of functional brain networks. Among multiple nodes that constitute the brain’s networks, several “social brain” regions have been proposed as key structures that particularly contribute to functional alterations in the ASD brain. However, because current standard analyses of functional connectivity (FC) are mainly driven by hypotheses related to deficits in interpersonal functions in ASD, highlight of abnormalities in the social brain may be an unsurprising consequence.

Objectives: In order to overcome potential biases of the present approaches, we adopted hypothesis-free data-driven analyses to a large cohort data of the resting-state FCs of adults with ASD collected at multiple sites and tested whether the importance of deficits in the social brain regions in ASD is supported without any hypotheses related to social functions.

Methods:  We collected the resting-state FC data of 74 adult high-functioning ASDs and 114 typically-developed controls collected at three different MRI sites in Japan. From each participant, we generated a correlation matrix consisted of 9,730 FCs using a standard brain atlas. Using this set of the correlation matrices as input, we combined advanced statistical methods to develop a classifier consisted of a small number of FCs that distinguished between ASD and TD. We then examined the anatomical distribution of the regions included in the selected FCs of the classifier.

Results: We developed a classifier that incorporated as little as 0.16% of all the 9,730 FCs in the brain (16 FCs). This classifier turned out to distinguish between ASD and TD with high precision (85 %) for the data collected at the three Japanese sites. Furthermore, we observed the generalization of this classifier into the third cohort collected in the USA sites with 75% accuracy. Anatomical examination of the 16 FCs revealed a right-sided distribution (31% right intra-hemispheric FCs as opposed to absent left intra-hemispheric FC). Furthermore, the 32 brain regions comprising the 16 FCs contained most of the important nodes in the social brain network, including the right amygdala, superior temporal sulcus, cingulate cortex, orbitofrontal cortex, and inferior frontal cortex.

Conclusions: The results suggest that abnormalities of a small number of FCs underlie core ASD deficits. Our data-driven approaches highlight the importance of the social brain FCs in the ASD brain and provide complementary findings to standard hypothesis-driven approaches.