Probing Autism and ADHD subtypes using cortical signatures of the T1w/T2w-ratio and morphometry

Autism spectrum disorder (ASD) and attention-deficit/hyperactivity disorder (ADHD) are neurodevelopmental conditions that share genetic etiology and frequently co-occur. Given this comorbidity and well-established clinical heterogeneity, identifying individuals with similar brain signatures may be valuable for predicting clinical outcomes and tailoring treatment strategies. Cortical myelination is a prominent developmental process, and its disruption is a candidate mechanism for both disorders. Yet, no studies have attempted to identify subtypes using T1w/T2w-ratio, a magnetic resonance imaging (MRI) based proxy for intracortical myelin. Moreover, cortical variability arises from numerous biological pathways, and multimodal approaches can integrate cortical metrics into a single network. We analyzed data from 310 individuals aged 2.6–23.6 years, obtained from the Province of Ontario Neurodevelopmental (POND) Network consisting of individuals diagnosed with ASD (n = 136), ADHD (n = 100), and typically developing (TD) individuals (n = 74). We first tested for differences in T1w/T2w-ratio between diagnostic categories and controls. We then performed unimodal (T1w/T2w-ratio) and multimodal (T1w/T2w-ratio, cortical thickness, and surface area) spectral clustering to identify diagnostic-blind subgroups. Linear models revealed no statistically significant case-control differences in T1w/T2w-ratio. Unimodal clustering mostly isolated single individual- or minority clusters, driven by image quality and intensity outliers. Multimodal clustering suggested three distinct subgroups, which transcended diagnostic boundaries, showing separate cortical patterns but similar clinical and cognitive profiles. T1w/T2w-ratio features were the most relevant for demarcation, followed by surface area. While our analysis revealed no significant case-control differences, multimodal clustering incorporating the T1w/T2w-ratio among cortical features holds promise for identifying biologically similar subsets of individuals with neurodevelopmental conditions.