Machine Learning-driven ADHD Classification: Exploring Medication Effects with VMD Sub-band Analysis

There has been increasing interest in neuroimaging studies in recent years, and computer-aided approaches have gained prominence in improving diagnostic accuracy. Attention Deficit Hyperactivity Disorder (ADHD) is a prevalent neurodevelopmental disorder characterized by inattention, impulsivity, and hyperactivity. Traditional diagnostic approaches often rely on subjective assessments, highlighting the need for more objective, datadriven methods. This study aims to classify ADHD subtypes and assess medication effects by converting resting-state fMRI images into one-dimensional (1D) signals and extracting statistical features using Variational Mode Decomposition (VMD).

Resting-state fMRI data from the ADHD-200 dataset, including 41 healthy controls (HC), 41 medicated ADHD-Combined (ADHD-C) individuals, and 41 non-medicated ADHD-C individuals, were analyzed. The 1D fMRI signals were decomposed into nine sub-bands using VMD. Statistical features were extracted from each sub-band and classified using Support Vector Machines (SVM), Linear Discriminant Analysis (LDA), and Artificial Neural Networks (ANN).

VMD-derived features substantially improved classification performance. The highest binary classification accuracy was achieved by LDA: 96.34% distinguishing non-medicated ADHD from controls and 88.41% for medicated ADHD versus controls. The classification between medicated and non-medicated ADHD yielded 79.63% accuracy. Ternary classification across all groups reached 69.51% accuracy.

These findings show that the VMD-based approach improves the classification of ADHD subtypes and helps evaluate medication effects. However, the lower performance in multi-class classification reflects the complexity of ADHD neuroimaging data.

The VMD-based approach improves classification accuracy, especially in distinguishing ADHD subtypes and medication effects, supporting its potential as an objective tool for diagnosis and treatment planning.