Discriminating ADHD From Healthy Controls Using a Novel Feature Selection Method Based on Relative Importance and...

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Attention-deficit/hyperactivity disorder (ADHD) is a neurodevelopmental disorder impacting children and adults, often diagnosed via clinical symptoms.
• Objective imaging biomarkers are needed for ADHD diagnosis due to the limitations of subjective clinical assessments.
• High-dimensional neuroimaging data with limited sample sizes present challenges in classifying brain disorders like ADHD.

Purpose of the Study:

• To develop and validate a novel feature selection method (FS_RIEL) for classifying ADHD using functional connectivity (FC) patterns from resting-state functional magnetic resonance imaging (rs-fMRI).
• To improve the accuracy of ADHD classification in both children and adults by addressing the challenge of high-dimensional data with limited samples.

Main Methods:

• Utilized resting-state functional magnetic resonance imaging (rs-fMRI) data from individuals with ADHD and age-matched healthy controls (HCs).
• Developed and applied a novel Feature Selection method based on Relative Importance and Ensemble Learning (FS_RIEL) to reduce feature dimensions.
• Compared the performance of FS_RIEL against traditional feature selection methods for ADHD classification.

Main Results:

• The FS_RIEL algorithm significantly improved ADHD classification accuracy by approximately 15% compared to traditional methods, achieving 80-86% accuracy.
• Identified frequently selected functional connectivities (FCs) primarily within the frontoparietal, default, salience, basal ganglia, and cerebellum networks.
• These findings suggest a widespread brain connectivity impairment profile in ADHD.

Conclusions:

• The FS_RIEL method offers an effective approach for ADHD classification using rs-fMRI data, particularly in overcoming high-dimensional feature challenges.
• The identified brain networks associated with ADHD may serve as potential objective biomarkers for early diagnosis.
• ADHD is characterized by a broad disruption in brain connectivity networks, impacting multiple functional systems.