Classification of ADHD subgroup with recursive feature elimination for structural brain MRI

Area of Science:

• Neuroscience
• Machine Learning
• Medical Imaging

Background:

• Attention Deficit Hyperactivity Disorder (ADHD) is a complex neurodevelopmental disorder with distinct subgroups.
• Accurate classification of ADHD subtypes is crucial for targeted treatment strategies.
• Existing diagnostic methods often lack the precision required for subgroup differentiation.

Purpose of the Study:

• To develop and validate a modified feature selection approach for binary classification of ADHD subgroups.
• To compare the efficacy of the proposed method against the standard RFE-SVM model.
• To identify significant neuroanatomical regions that could serve as biomarkers for ADHD classification.

Main Methods:

• Utilized the ADHD-200 dataset for analysis.
• Implemented a modified Recursive Feature Elimination Support Vector Machine (RFE-SVM) model for feature selection.
• Performed ten-fold cross-validation to evaluate classification performance.
• Compared performance metrics including J-statistics, F1-score, and classification accuracy.

Main Results:

• The modified RFE-SVM approach demonstrated superior performance compared to the original RFE-SVM.
• Achieved a classification accuracy of 84.17% using a linear Support Vector Machine (SVM) classifier.
• Identified specific anatomical regions with significant potential as biomarkers for ADHD subgroup classification.

Conclusions:

• The proposed modified RFE-SVM feature selection method is effective for ADHD subgroup classification.
• Neuroanatomical features hold promise as reliable biomarkers for differentiating ADHD subtypes.
• This approach can aid in developing more precise diagnostic and therapeutic strategies for ADHD.