Multivariate classification of autism spectrum disorder using frequency-specific resting-state functional connectivity—A multi-center study

• Published in: Progress in neuro-psychopharmacology & biological psychiatry Vol. 64; pp. 1 - 9
• Main Authors: Chen, Heng, Duan, Xujun, Liu, Feng, Lu, Fengmei, Ma, Xujing, Zhang, Youxue, Uddin, Lucina Q., Chen, Huafu
• Format: Journal Article
• Language: English
• Published: England 04.01.2016
• Subjects: Adolescent; Autism Spectrum Disorder - classification; Autism Spectrum Disorder - physiopathology; Brain - physiopathology; Brain Mapping; Child; Data Interpretation, Statistical; Databases, Factual; Female; Head Movements; Humans; Magnetic Resonance Imaging - methods; Male; Multivariate Analysis; Neural Pathways - physiopathology; Rest; Severity of Illness Index; Support Vector Machine; Functional connectivity; Slow-4; Multivariate pattern analysis; Autism spectrum disorder
• DOI: 10.1016/j.pnpbp.2015.06.014
• ISSN: 0278-5846; 1878-4216

Resting-state functional magnetic resonance imaging studies examining low frequency fluctuations (0.01-0.08 Hz) have revealed atypical whole brain functional connectivity patterns in adolescents with autism spectrum disorder (ASD), and these atypical patterns can be used to discriminate individuals with ASD from controls. However, at present it is unknown whether functional connectivity at specific frequency bands can be used to discriminate individuals with ASD from controls, and whether relationships with symptom severity are stronger in specific frequency bands. We selected 240 adolescent subjects (12-18 years old, 112 with autism spectrum disorder (101/11, males/females) and 128 healthy controls (104/24, males/females)) from 6 separate international sites in the Autism Brain Imaging Data Exchange database. Whole brain functional connectivity networks were constructed in the Slow-5 (0.01-0.027 Hz) and Slow-4 (0.027-0.073 Hz) frequency bands, which were then used as classification features. An accuracy of 79.17% (p<0.001) was obtained using support vector machine. Most of the discriminative features were concentrated on the Slow-4 band. In the Slow-4 band, atypical connections between the default mode network, fronto-parietal network and cingulo-opercular network were detected. A significant correlation was found between social and communication deficits as measured by the ADOS in individuals with ASD and the classification scores based on connectivity between the default mode network and the cingulo-opercular network. Connections of the thalamus were of the highest classification weight in the Slow-4 band. Our findings provide preliminary evidence for frequency-specific whole brain functional connectivity indices that may eventually be used to aid detection of ASD.