Using a self-organizing map algorithm to detect age-related changes in functional connectivity during rest in autism spectrum disorders

Abstract

Healthy individuals show robust functional connectivity during rest, which is stronger in adults than in children. Connectivity occurs between the posterior and anterior portions of the default network, a group of structures active in the absence of a task, including the posterior cingulate cortex and the superior frontal gyrus. Previous studies found weaker posterior–anterior connectivity in the default network in adults and adolescents with autism spectrum disorders (ASD). However, these studies used small a priori regions of interest (“seeds”) to calculate connectivity. Since seed location for all participants was chosen based on controls' brains, these studies' analyses are more tailored to controls than individuals with ASD. An alternative is to use a data-driven approach, such as self-organizing maps (SOM), to create a reference for each participant to calculate connectivity. We used individualized resting-state clusters identified by an SOM algorithm to corroborate previous findings of weaker posterior–anterior connectivity in the ASD group and examine age-related changes in the ASD and control groups. Thirty-nine adolescents with ASD and 41 controls underwent a 10-minute, eyes-open, resting-state functional MRI scan. The SOM analysis revealed that adolescents with ASD versus controls have weaker connectivity between the posterior hub of the default network and the right superior frontal gyrus. Additionally, controls have larger increases in connectivity with age compared to the ASD group. These findings indicate that SOM is a complementary method for calculating connectivity in a clinical population. Additionally, adolescents with ASD have a different developmental trajectory of the default network compared to controls.