Anatomical Connectivity-Based Analysis of Autism Using Diffusion Tensor Imaging

Background: Studies showed that Autism Spectrum Disorder (ASD) might result from abnormal connections among different anatomical regions rather than defects in specific brain area. Using Diffusion Tensor Imaging (DTI), inter-regional connectivity can be defined quantitatively by the white matter fiber tracts.

Objectives: To define inter-regional connectivity using DTI and compare them between high-functioning ASD and typically developing (TD) controls. 

Methods: Ten high functioning autistics and 10 typically-developing (TD) matched subjects were studied.To define inter-regional connectivity, first, an elastic registration algorithm was recruited to align the JHU-DTI-MNI atlas onto each subject to automatically label the brain regions. Then, an improved tensor-based fast marching  method was employed to simulate water diffusion dynamics to define the connectivity strengths among different regions. Specifically, the connectivity strength between two regions is defined by a combination of the fast marching results, “time map” and “velocity map”, which starts the diffusion simulation from one region and terminate to another. Faster diffusion between these two regions indicates denser, highly oriented fibers, or stronger connectivity between them. Finally, the connectivity strengths among 46 selected anatomical regions from the cortical surface were quantified, and statistics analysis was performed to study the connectivity strength of autistics by comparing with that of controls. 

Results: Statistical analysis on the connectivity strength between each anatomical region pair showed that the regions with a large number of significantly different stronger connections in ASD compared with TD (p-value<0.05) include the left and right superior occipital gyri, supramarginal gyri, and the left middle occipital gyrus, angular and pre-cuneus.  Among the 67 significantly stronger connections (p-value<0.05), the smallest p-values were from the connections between cuneus (R) and middle occipital gyrus (R), pre-cuneus (L) and supramarginal gyrus (L), inferior frontal gyrus (L) and superior parietal lobule (L), inferior frontal gyrus (R) and superior parietal lobule (R) as well as superior parietal lobule (L) and supramarginal gyrus (L). On the other hand, the regions with a large number of significant stronger connections in TD compared to ASD include the left inferior temporal gyrus, inferior frontal gyrus, middle frontal gyrus, insular and lateral fronto-orbital gyrus. Among the 16 significantly different connections, the top lists are those between inferior temporal gyrus (L) and lateral fronto-orbital gyrus (L), inferior temporal gyrus (L) and middle fronto-orbital gyrus (L), inferior temporal gyrus (L) and precentral gyrus (L), inferior temporal gyrus (L) and insular, as well as inferior frontal gyrus (L) and supramarginal gyrus (L). After reordering the average connectivity matrix and permuting rows and columns so that highly connected regions are rearranged as neighboring index, three local clusters can be clearly seen from the connectivity matrix, and the majority of the above significantly different connections are within these clusters, indicating local stronger connectivity within left and right occipital, temporal, parietal lobes. 

Conclusions: A novel inter-region connectivity quantification algorithm was proposed, and statistical analysis showed that high-functioning ASD have more local connectivity among brain regions as compared to TD, and significantly different connections were found in visual and multimodal integration regions.