Can Resting-State fMRI be Used to Inform Differential Diagnosis between ASD and ADHD?

Background: Autism spectrum disorder (ASD) and attention deficit/hyperactivity disorder (ADHD) are difficult to differentiate in clinical settings (Miodovnik et al., 2015; Smith et al., 2017), as both exhibit atypical patterns of attention and executive functioning (Gioia et al., 2002; Johnson et al., 2014). Further, children with ASD and ADHD have shared and distinctive patterns of long-range functional connectivity in the brain (Di Martino et al., 2013), though such patterns have not yet been used to mechanistically distinguish between these two disorders. Two brain networks have been implicated in these disorders and are of interest: the salience network (SN), which is related to attentional processes for internal and external events, and the FPN, which plays a role in flexible goal-driven behavior (Seeley et al., 2007). Functional connectivity between these two networks may reveal meaningful subgroups for those with ASD vs. ADHD.

Objectives: The aim of this project is to determine whether community detection, a graph-theory algorithm for subtype identification, can differentiate children with ASD from children ADHD using patterns of SN and FPN connectivity.

Methods: Functional and structural Magnetic Resonance Imaging (MRI), and phenotypic data were selected from the Autism Brain Imaging Data Exchange (ABIDE), its follow-up project ABIDE-2, and the ADHD-200 database. All data were used from New York University Langone Medical Center in order to match datasets on scanner protocol. After excluding subjects with high motion the final dataset consisted of 60 children with ASD, 64 children with ADHD, and 65 typically developing (TD) controls. 5mm spherical regions of interest (ROIs) were chosen for the SN and FPN using Neurosynth (Yarkoni et al., 2011). Group Iterative Multiple Model Estimation (GIMME), a graph theory approach (Gates and Molenaar, 2012), was used with scrubbed resting state data from each subject. Subsequently, a community detection algorithm was used to identify subgroups characterized by SN-FPN connectivity patterns.

Results: ASD and ADHD groups did not differ on age, FSIQ, sex, and motion (all ps>.21). GIMME identified two subgroups. Subgroup A had 70 cases and was characterized by hyperconnectivity of SN-FPN nodes. Subgroup B had 119 cases and was characterized by hypoconnectivity of SN-FPN nodes. A chi-square test was performed and a relationship was found between diagnostic status (ASD, ADHD, TD) and the frequency of subgroup membership, X² (1, N=189)=7.18, p<.05. Of note, 71.7% of the ASD group and 67.7% of the TD group were in subgroup B. The ADHD group demonstrated more heterogeneity in network patterns with 50% in subgroup B.

Conclusions: ASD and ADHD were not characterized by distinct subgroup patterns, though analyses revealed more heterogeneity in ADHD than in ASD. Hyperconnectivity in SN-FPN nodes may reveal specific inattention or executive function difficulties. These results suggest potential for functional connectivity of the SN-FPN in understanding the similarities and differences in ASD vs. ADHD. It will be important for future research to collect phenotypic data in symptom severity and executive function domains in order to best understand what factors impact subgroup membership and play a role in the heterogeneity of brain networks.