International Meeting for Autism Research: Brain Network-Based Analysis of Autism Using Diffusion Tensor Imaging

Brain Network-Based Analysis of Autism Using Diffusion Tensor Imaging

Saturday, May 14, 2011: 3:00 PM
Elizabeth Ballroom D (Manchester Grand Hyatt)
1:15 PM
H. Li1, Z. Xue1, T. M. Ellmore2, R. E. Frye2, B. Malmberg2 and S. T. Wong1, (1)Bioengineering and Bioinformatics Program, The Methodist Hospital Research Institute, Weill Cornell Medical College, Houston, TX, (2)University of Texas Houston Health Science Center, Houston, TX
Background:  In addition to connectivity among different brain regions, brain network topology acts as an important characteristic for Autism Spectrum Disorder (ASD). Previous studies have suggested that ASD can be characterized by increased local and decreased long-range connectivity. As a new MR modality to track white matter fiber bundles, Diffusion Tensor Imaging (DTI) can be effectively used to quantify the structural connectivity network of human brains, thereby define neuroimaging biomarkers using network-based analysis for ASD.

Objectives: To define brain connectivity network and compare the network characteristics between high-functioning ASD and typically developing (TD) controls.

Methods: Ten high functioning autistics (aged 7-14) and 10 typically-developing matched subjects were studied. All subjects underwent a single MR imaging session at 3T (Philips Intera), including T1-weighted MRI (sagittal, TR/TE=8.4/3.9ms;FA=8 degrees; slice thickness = 1.0 mm) and a 32-direction diffusion imaging sequence (high angular resolution, overplus on, TR/TE=8500/67 ms; FA=90 degrees; 2 mm axial slice thickness, max. b-value of 800 s/mm2).After applying eddy current correction, tensor calculation and tissue segmentation, an elastic registration was utilized to align the JHU-DTI-MNI atlas onto each subject to automatically label the brains. Then, an improved tensor-based fast marching method was employed to simulate water diffusion dynamics to define the connectivity strengths among different neuroanatomical regions. According to our algorithm, the marching speed is faster for higher diffusion, thus the result indicates denser, highly oriented fibers, or stronger connectivity between different regions. Finally, the connectivity strengths among 46 selected anatomical regions overlapping the cortical surface were quantified to construct the brain neuroanatomical connectivity network and the network-based analysis. The Clustering coefficient (CC) and characteristic path length (CL) for the entire networks were quantified for each group. Further, the CC and CL of each sub-network associated with each anatomical region were also calculated. 

Results: Statistical analysis showed that autistics have higher CC (0.18±0.01 vs. 0.16±0.01, two sample single-tailed t-test, p-value=0.002) and lower CL (1.71±0.05 vs. 1.75±0.04, p-value=0.03) compared with controls, indicating stronger local connectivity and shorter network connection paths between regions in ASD as compared to TD controls. For sub-network analysis, the CC and CL of the sub-networks connected to each anatomical region are calculated, and the anatomical regions with significant different local network characteristics (higher CC and lower CL) included left and right supramarginal, superior occipital, middle occipital, angular gyri, and the left pre-cuneus. Interestingly, these regions are associated with visual and multimodal integration. 

Conclusions: The brain network-based analysis revealed that children with high-functioning ASD have stronger local connectivity among brain regions as compared to TD children. Significantly different local network characteristics were found in visual and multimodal integration regions. The method may open up a new vista to decode the mystery of ASD at the neuroanatomical level.

| More