Atomic dynamic functional interaction patterns for characterization of ADHD.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Psychiatric Disorders

Background:

• Modeling temporal dynamics of functional interactions is crucial for understanding psychiatric disorders.
• Previous research has rarely explored disease-related abnormalities within data-driven subnetworks.
• Attention-deficit/hyperactivity disorder (ADHD) presents complex challenges in understanding functional brain dynamics.

Purpose of the Study:

• To propose a novel computational framework for simultaneous modeling of functional brain interactions and their dynamics.
• To assess functional interaction abnormalities within data-driven subnetworks in ADHD.
• To differentiate children with ADHD from normal controls (NC) using identified functional patterns.

Main Methods:

• Developed a Bayesian connectivity change point model for dynamic functional interactions.
• Employed a nonnegative matrix factorization variant to identify abnormal subnetworks.
• Applied the framework to resting-state fMRI data from children with ADHD and NC.

Main Results:

• Identified four atomic functional interaction patterns (AFIPs): two for ADHD and two for NC.
• Grouped AFIPs into common and abnormal pairs, highlighting ADHD-specific patterns.
• Derived two data-driven abnormal functional subnetworks based on the abnormal AFIP pair.
• Achieved successful differentiation of all ADHD children from NCs based on AFIP approximation.

Conclusions:

• The proposed framework effectively models functional brain interaction dynamics and identifies abnormalities in ADHD.
• The identified abnormal subnetworks are significant indicators for ADHD diagnosis.
• This approach demonstrates high potential for accurate diagnostic differentiation in psychiatric neuroimaging studies.