Differentiating Boys with ADHD from Those with Typical Development Based on Whole-Brain Functional Connections Using...

Area of Science:

• Neuroimaging
• Machine Learning
• Neuroscience

Background:

• Machine learning (ML) shows promise in differentiating clinical populations from healthy individuals.
• Resting-state functional magnetic resonance neuroimaging (R-fMRI) utilizes interregional functional connections as discriminative features.
• Previous studies have explored ML for ADHD detection, but comprehensive whole-brain connectivity patterns require further investigation.

Purpose of the Study:

• To investigate spatially distributed, ADHD-related discriminative features from whole-brain resting-state functional connectivity (FC) patterns.
• To apply machine learning techniques to classify individuals with ADHD from typically developing controls based on R-fMRI data.
• To identify specific brain regions and connections critical for discriminating ADHD.

Main Methods:

• Acquired R-fMRI data from 40 individuals with ADHD and 28 healthy controls.
• Employed machine learning, specifically Support Vector Machine (SVM) with leave-one-out cross-validation (LOOV), for classification.
• Assessed classification performance using permutation tests.

Main Results:

• Achieved 85.3% classification accuracy in distinguishing ADHD patients from controls.
• Identified the cerebellum, Default Mode Network (DMN), and frontoparietal regions as key areas with discriminative FC.
• Found significant correlations between cerebellar-DMN functional connections and ADHD behavioral symptoms.

Conclusions:

• Whole-brain resting-state functional connections offer valuable neuroimaging biomarkers for ADHD.
• These findings suggest potential for ML-based neuroimaging in clinically assisting ADHD diagnosis.
• The cerebellum plays a crucial role in the functional connectivity alterations observed in ADHD.