Analysis of High Quality Diffusion Tensor Imaging in Young Children with Autism

Background: Diffusion tensor imaging (DTI) shows promise for studying potential structural abnormalities in the brains of autistic children. The most consistent finding in the literature is reduced fractional anisotropy (FA) and increased mean diffusivity (Trace(D)) in autistic subjects compared to healthy subjects. However, regional distribution of the abnormalities is inconsistent across studies.  Some of this heterogeneity may be due to small sample sizes, which might obscure meaningful between-group differences that fail to reach the threshold set for statistical significance.  Further, the magnitude of the reported changes is small, rendering them vulnerable to the effects of minor confounds. 

Objectives: To investigate the brains of autistic children as compared to age and gender matched typically developing children using high quality DTI data, and to create average brains of each group, reporting the regional magnitude of differences in various DTI metrics.

Methods: 39 children, 2.2 – 8.7 years, mean 4.6 ± 1.7 years, 28 male, who met DSM-IV criteria for autism and 39 age and gender matched typically developing children, 2.0 – 8.1 years, mean 4.7 ± 1.8 years, 26 male, were scanned on a 1.5T GE scanner.  DTI data consisted of 60 b=1100s/mm², 10 b=300s/mm² and 10 b=0s/mm² volumes at 2.5mm isotropic resolution.  Data was preprocessed using TORTOISE to correct for motion and distortions.  Tensor based registration (DTITK) was used to create an average brain for the population for voxelwise analysis of DTI metrics including FA, Trace(D), radial diffusivity (RD) and axial diffusivity (AD).  Average tensor derived metrics were computed for the autistic and typically developing children separately.  Directionally encoded color maps were visually assessed, and subtraction maps were computed for all metrics. Traditional TBSS analysis was also performed on all metrics.

Results: TBSS analysis showed a reduction in FA in autistic children compared to typically developing children in many white matter regions, including the cerebellum, genu, splenium and body of the corpus callosum, brain stem, posterior limb of the internal capsule, superior frontal and temporal parietal regions.  Trace(D) was greater in autistic compared to typically developing children in most white matter regions, but only in the posterior half of the brain. AD and RD results mimicked Trace(D).  Subtraction of average maps showed very small magnitude differences, e.g. FA difference in genu of the corpus callosum is less than 1%, and Trace(D) in the splenium of the corpus callosum is about 1%.

Conclusions: The general trend of decreased FA and increased Trace(D) in autistic subjects is consistent with previously reported studies.  However, the distribution of the abnormal regions adds to the heterogeneity of the existing inconsistencies in the literature.  Further, the magnitude of the differences is small and may reflect between-group differences that are not related to brain anatomy.  For example, the anterior-posterior gradient of Trace(D) we observed may be the result of a small difference in subject motion between groups.  While DTI is promising as a method for revealing anatomic abnormalities in autism, caution must be exercised in interpreting between-group differences.  Replication of published findings is a crucial first step.