Linking ADHD and Behavioral Assessment Through Identification of Shared Diagnostic Task-Based Functional Connections

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning

Background:

• Existing research suggests atypical brain connectivity in attentional, reward, and task inhibition networks is linked to Attention Deficit Hyperactivity Disorder (ADHD).
• The neural basis for how behavioral tasks aid in ADHD diagnosis remains unclear.

Purpose of the Study:

• To investigate if machine learning classifiers can use task-based functional connectivity to predict ADHD diagnosis and behavioral task performance.
• To identify specific connectivity signatures that link ADHD diagnosis to behavioral phenotypes.

Main Methods:

• Analyzed archival MRI and behavioral data from 80 participants (ADHD vs. Control) who completed a go/no-go task.
• Measured functional connectivity using cross-mutual information during task performance.
• Employed multilayer feedforward classifier models to identify predictive functional connections for diagnosis and Iowa Gambling Task (IGT) performance.

Main Results:

• Machine learning models accurately predicted clinical diagnosis (ADHD vs. Control) and IGT performance with 0.91 accuracy and high sensitivity/specificity (d' > 2.9).
• Key diagnostic functional connections were identified between visual, ventral attentional, and anterior default mode networks.
• Task-based functional connectivity demonstrated utility as a biomarker for ADHD.

Conclusions:

• Task-based functional connectivity serves as a reliable biomarker for ADHD.
• The developed analytical framework links behavioral assessments to both clinical diagnosis and functional connectivity.
• This approach may improve differential diagnosis and inform targeted intervention strategies for ADHD.