Resting-state brain scans found reduced directional connectivity in autistic children, with the strongest differences concentrated in networks involved in self-reflection and social cognition.
A study published in Frontiers in Psychiatry has found that autistic and typically developing children show different patterns of directed brain connectivity at rest, with the strongest changes appearing in the default mode network, a set of regions linked to self-referential thought, social cognition and internal mentation.
Using resting-state functional MRI data from the Autism Brain Imaging Data Exchange and focusing on a single site to reduce scanner-related variation, the researchers compared 52 children and adolescents with autism spectrum disorder with 52 matched controls. They applied convergent cross mapping, a nonlinear method designed to detect directional influence between brain regions, rather than relying only on conventional correlation measures. Across five resting-state networks, the team reported an overall reduction in causal connectivity in the autism group, alongside a smaller number of local increases.
The default mode network stood out most clearly. According to the study, it contained 13 altered directed connections in uncorrected analyses, and six remained significant after false discovery rate adjustment. The largest differences involved frontal and cingulate regions, including medial superior frontal, orbitofrontal and anterior cingulate areas, as well as the posterior cingulate and inferior temporal cortex. The attention network, sensorimotor network, subcortical network and visual network also showed changes, but fewer survived correction for multiple testing. The authors said the pattern points to widespread but uneven disruption in network communication, with some strengthened links possibly reflecting compensation, although that interpretation remains tentative.
To test whether these connectivity patterns could distinguish the two groups, the researchers trained a support vector machine using the causal connectivity features. The model achieved 82% accuracy, with sensitivity of 83%, specificity of 81% and an area under the curve of 0.89. The most important features were concentrated in the default mode network, reinforcing the idea that this system may be especially vulnerable in autism. The authors said the findings could help refine understanding of the disorder’s neurobiology, but they also stressed the limits of the work, including the modest sample size, the single-site design and the need for replication in larger, multi-centre datasets.
Disclaimer: This content is for informational purposes only and is not intended to be a substitute for professional medical judgment, advice, diagnosis, or treatment.
