A study using structural MRI found clusters of brain-structure connections that differed between autistic young people and non-autistic peers of the same sex. The findings appeared in both boys and girls, but a direct statistical test did not find a significant autism-related difference between the two groups.
The research, published in Cerebral Cortex, included 337 participants aged 8 to 18. The group consisted of 181 autistic young people—97 boys and 84 girls—and 156 non-autistic young people—78 boys and 78 girls. Participants were scanned at four US research sites as part of the first wave of the NIH Autism Centers of Excellence Network.
Different network hubs appeared in boys and girls
Among autistic boys compared with non-autistic boys, 35 linked pairs of brain regions differed. Most showed regions that were more similar in physical structure in the autistic group. The cluster was led by the left fusiform gyrus and also included the left medial orbitofrontal cortex, right entorhinal cortex and right lingual gyrus.
Among autistic girls compared with non-autistic girls, 41 connections differed, mostly in the same direction. The main hub was the entorhinal cortex on both sides of the brain. The network extended to the inferior parietal lobule, lateral occipital cortex and areas involved in movement and touch. A smaller set of decreased connections mainly involved the superior temporal sulcus, including one connection with the fusiform gyrus.
These clusters should not be read as evidence that autism has one brain pattern in boys and another in girls. When the researchers directly tested whether autism-related patterns differed by sex, they found no significant cluster at any of the three thresholds examined. They said the apparent contrast between the clusters could reflect noise, or that any genuine difference may be too small for this sample to detect.
What the MRI method measured
The scans showed brain anatomy rather than brain activity. Researchers measured cortical thickness, surface area, sulcal depth, mean curvature and gray-matter volume across 68 cortical regions. They then used MIND, or Morphometric Inverse Divergence, to compare the physical similarity of every pair of regions, producing 2,278 region-to-region connections for each participant.
This approach measures relationships in brain morphology. It does not directly measure nerve fibers, white-matter wiring or moment-to-moment brain activity. Network-based statistical analysis looked for connected groups of regions that differed between groups after accounting for age, head size and scan site.
Links with behavior were limited
In boys, the overall autism-related network was associated with social and communication scores from the Autism Diagnostic Interview-Revised. The correlation was modest, at r = 0.39, and researchers described it as negative: stronger connectivity was associated with less severe social-communication scores.
The corresponding network in girls was not significantly linked to autism trait scores. Neither sex’s network was associated with parent-checklist scores for anxiety, mood, attention or behavior problems.
The researchers also identified a separate 48-connection network that distinguished non-autistic boys from non-autistic girls. It shared no connections with the boys’ autism-related cluster and one connection with the girls’ cluster.
Why the findings are not a diagnostic test
The network strength separated autistic from non-autistic participants within this sample with an area under the curve of 0.72 in boys and 0.66 in girls. However, the same participants were used to identify the networks and calculate these figures. They were not tested in a separate group, so the performance may be inflated.
Removing cortical thickness reduced the area under the curve to 0.54 in boys and 0.58 in girls, while removing surface area or volume made little difference. The researchers said the approach does not measure diagnostic usefulness and called for machine-learning studies with outside testing.
The results are group-level averages with substantial overlap. Nothing in the study shows whether a single scan can establish that a child is autistic. The cross-sectional design also cannot show how the patterns develop over time. Because the participants were 8 to 18, the findings do not describe brain-structure changes across adulthood.
Most participants were White and were recruited through four academic centers. The study used structural MRI only, and both clusters remained after adjustment for scan quality and scan site. Further research will be needed to determine how reproducible these networks are across ages, populations and independent samples.
AI tools were used to assist with the preparation of this article under our editorial standards.
Leave a Reply