Altered Resting State Functional Network Topology Across Neurological Conditions: A Comparison of Autism Spectrum Disorder with Phenylketonuria and Traumatic Brain Injury
Objectives: In the present study, we used graph theoretical analysis to examine resting state FC data from a large sample of participants in hopes of gaining additional insight into how the topological properties of functional networks differ based on factors such as diagnosis and age. In addition to individuals with ASD and typically developing individuals (TD), we examined individuals with phenylketonuria (PKU), a metabolic disorder with cognitive symptoms, and individuals with traumatic brain injury (TBI).
Methods: Resting state fMRI data was collected from 61 individuals with ASD (mean age=15.4), 12 individuals with PKU (mean age=23.6), 18 individuals with TBI (mean age=39.4), and a comparison group of 61 TD individuals (mean age=15.4). The ASD and TD groups were also split into age subgroups via median split. Following data pre-processing and anatomical parcellation into 90 cortical and subcortial regions of interest, partial correlation matrices were generated and thresholded at consistent network sparsity. Topological properties were then compared between diagnostic groups and age subgroups via two- and one-way ANOVAs.
Results: Statistical analysis revealed that the ASD group demonstrated reduced local network organization (decreased local network efficiency and likelihood of short-range connections, p < .05 in both instances) that accompanied a bias toward greater global network organization (increased global network efficiency and likelihood of long-range connections, p < .05 in both instances), as compared to the TD group. There were no significant interactions between diagnosis and age groups. The PKU and TBI groups demonstrated a bias toward local network organization (increased local network efficiency and likelihood of short-range connections, p < .001 in both instances) with reduced global network organization (decreased global network efficiency and likelihood of long-range connections, p < .001 in both instances), in comparison to the TD group. In subsequent analyses, 13 functional subnetworks were identified using a community detection algorithm and compared between groups for topology. In particular, as compared to the TD group, the ASD group demonstrated reduced network density within a subnetwork containing temporal cortical regions (p < .01). The PKU group showed reduced FC in subnetworks containing frontoparietal and orbitofrontal regions, as compared to the TD group (p < .01). Lastly, the TBI group displayed widespread reduced network efficiency and FC in 10 of the 13 subnetworks (p < .01 in all instances).
Conclusions: The results of the present study indicate alterations in functional network and subnetwork topology in ASD. In addition, ASD seems to differ from other neurological conditions in its bias toward global over local network organization. Future studies are needed to explore how these findings contrast with other FC findings in ASD, as well as characterize network alterations within the context of symptom severity.