Connectivity-based parcellation increases network detection sensitivity in resting state fMRI: An investigation into...

Although resting state fMRI (RS-fMRI) is increasingly used to generate biomarkers of psychiatric illnesses, analytical choices such as seed size and placement can lead to variable findings. Seed placement especially impacts on RS-fMRI studies of Autism Spectrum Disorder (ASD), because individuals with ASD are known to possess more variable network topographies. Here, we present a novel pipeline for analysing RS-fMRI in ASD using the cingulate cortex as an exemplar anatomical region of interest. Rather than using seeds based on previous literature, or gross morphology, we used a combination of structural information, task-independent (RS-fMRI) and task-dependent functional connectivity (Meta-Analytic Connectivity Modeling) to partition the cingulate cortex into six subregions with unique connectivity fingerprints and diverse behavioural profiles. This parcellation was consistent between groups and highly replicable across individuals (up to 93% detection) suggesting that the organisation of cortico-cingulo connections is highly similar between groups. However, our results showed an age-related increase in connectivity between the anterior middle cingulate cortex and right lateral prefrontal cortex in ASD, whilst this connectivity decreased in controls. There was also a Group × Grey Matter (GM) interaction, showing increased connectivity between the anterior cingulate cortex and the rectal gyrus in concert with increasing rectal gyrus GM in controls. By comparing our approach to previously established methods we revealed that our approach improves network detection in both groups, and that the ability to detect group differences using 4 mm radius spheres varies greatly with seed placement. Using our multi-modal approach we find disrupted cortico-cingulo circuits that, based on task-dependent information, may contribute to ASD deficits in attention and social interaction. Moreover, we highlight how more sensitive approaches to RS-fMRI are crucial for establishing robust and reproducible connectivity-based biomarkers in psychiatric disorders.