Classifying Attention Deficit Hyperactivity Disorder (ADHD) through MRI images by Fuzzy-C-means and Advanced Linear Discriminant Analysis

Abstract

Attention Deficit Hyperactivity Disorder (ADHD) is one of the most prevalent neurodevelopmental disorders among children worldwide. Conventional diagnosis is primarily based on behavioral and clinical symptom evaluation, which may introduce subjectivity and reduce diagnostic reliability. Although Magnetic Resonance Imaging (MRI) has shown promise in identifying structural brain abnormalities linked to ADHD, existing computational approaches frequently encounter challenges such as high-dimensional feature spaces, inefficient segmentation, suboptimal feature discrimination, classifier overfitting, and limited generalization across heterogeneous datasets. To overcome these limitations, this research proposes an integrated FCM–ALDA framework designed to enhance segmentation accuracy and classification robustness. MRI data obtained from the ADHD-200 Consortium are utilized for model development and validation. The proposed methodology initially performs binary classification to distinguish ADHD subjects from typically developing controls. Subsequently, confirmed ADHD cases are further categorized into three clinical subtypes: ADHD-Inattentive (ADHD-I), ADHD-Hyperactive/Impulsive (ADHD-HI), and ADHD-Combined (ADHD-C). Feature extraction is centered on the Caudate Nucleus, a region critically associated with attention regulation and executive functioning. The Fuzzy C-Means (FCM) clustering algorithm is employed to segment MRI images and extract texture-, intensity-, and shape-based dynamic features, forming the Caudate Nucleus Dataset (CNDS). These features are then classified using Advanced Linear Discriminant Analysis (ALDA) to maximize inter-class separability while minimizing intra-class variance. Comparative evaluation against K-Nearest Neighbor (KNN) and Binary-Coded Genetic Algorithm optimized Extreme Learning Machine (BCGA-ELM) demonstrates that the proposed model achieves superior Accuracy, Sensitivity, and Specificity, confirming its effectiveness for MRI-based ADHD subtype classification.