20218
Different Patterns of Cortical Brain Alterations in Preschool-Aged Boys with Autism Spectrum Disorder with and without Intellectual Disability

Saturday, May 16, 2015: 11:30 AM-1:30 PM
Imperial Ballroom (Grand America Hotel)
Y. Feng1, H. Ota2,3, S. J. Rogers3, D. G. Amaral3, F. Hoeft4 and C. W. Nordahl3, (1)University of California Davis Medical Center, Health Informatics Program, Sacramento, CA, (2)Psychiatry, Showa University School of Medicine, Tokyo, Japan, (3)MIND Institute and Department of Psychiatry and Behavioral Sciences, University of California Davis Medical Center, Sacramento, CA, (4)Psychiatry, University of California at San Francisco, San Francisco, CA
Background:  Autism spectrum disorder (ASD) is a heterogeneous condition, with varying behavioral presentations and comorbidities. Multiple etiologies and neural phenotypes likely exist. An estimated 30% of children diagnosed with ASD have comorbid intellectual disability (ID). Little is known, however, about how the neuroanatomical profiles differ between children with ASD with and without comorbid ID.

Objectives:  We utilized multivariate pattern analyses to compare regional cortical gray matter measurements in preschool-aged boys divided into three groups: (1) ASD and comorbid ID (ASD+ID), (2) ASD without ID (ASD), and (3) age-matched typical developing (TD) controls.

Methods:  Structural MRIs were acquired in 105 boys (28 ASD, 27 ASD+ID, 50 TD) with a mean age of 36.2 months. Cortical gray matter was parcellated into 34 gyral regions per hemisphere using Freesurfer (v5.1.0). Measurements included surface area, cortical thickness, and volume for each cortical region. Cognitive testing was carried out at two time points, one at study entry and again two years later (mean 65.9 months) using the Mullen Scales of Early Development. Development Quotient (DQ) scores were calculated at both time points, and a cutoff of 70 was used to determine ASD+ID and ASD groups. Only participants with stable DQ scores across both time points were included in this study. We utilized cross-validated linear support vector machine (SVM) analyses, controlling for total cerebral volume, to classify ASD+ID vs TD, ASD vs TD, and ASD+ID vs ASD. Left and right hemispheres were analyzed separately. Preprocessing included normalization across features. Recursive feature elimination was utilized to identify features with the greatest contribution to classifications. Result includes: (1) performance (accuracy, sensitivity, specificity) (2) weights of regions that contribute to each classification, and (3) comparison of features across three classification groupings.

Results:  Classification accuracies were significant for all three comparisons. Left cortical volumes had the best discriminability for both ASD groups relative to TD, and there was significant overlap in the pattern of brain regions, particularly in the frontal and temporal lobes (e.g. temporal pole, medial orbitofrontal cortex, frontal pole, pars triangularis, caudal middle frontal gyrus, and pars orbitalis). There were, however, an additional 15 regions identified in the ASD+ID vs TD comparison diffusely distributed the entire brain. In contrast, there was only one region (entorhinal cortex) that was unique to the ASD vs TD comparison. Classification performance for discriminating between the two ASD subgroups (ASD+ID vs ASD) was also high. While cortical thickness did not achieve high accuracy in discriminating both ASD groups from TD, right cortical thickness had the highest discriminability for ASD +ID vs. ASD.

Conclusions:  Children with ASD with and without comorbid ID have overlapping, but different neural phenotypes. Children with ASD+ID have a more diffuse pattern of alterations relative to TD controls. A better characterization of unique patterns of brain alterations in different subgroups of children with ASD may lead to more specific, targeted intervention strategies.