Characterization and Classification of ADHD Subtypes: An Approach Based on the Nodal Distribution of Eigenvector Centrality and Classification Tree Model.

Abstract

In recent times, the complex network theory is increasingly applied to characterize, classify, and diagnose a broad spectrum of neuropathological conditions, including attention deficit hyperactivity disorder (ADHD), Alzheimer's disease, bipolar disorder, and many others. Nevertheless, the diagnosis and associated subtype identification majorly rely on the baseline correlation matrix obtained from the functional MRI scan. Thus, the existing protocols are either full of personalized bias or computationally expensive as network complexity-based simple but deterministic protocols are yet to be developed and formalized. This article proposes a deterministic method to identify and differentiate the common ADHD subtypes, which is based on a single complexity measure, namely the eigenvector centrality. The node-wise centrality differences were explored using a classification tree model (p < 0.05) to diagnose the subtypes. Identification of marker nodes from default mode, visual, frontoparietal, limbic, and cerebellar networks strongly vouch for the involvement of multiple brain regions in ADHD neuropathology.