Deep Learning-Based Classification of ADHD: A Study on the ADHD-200 Dataset
Bildiriler > Tam Metin Bildiri · 2025
Yazarlar: Selin Şaşal, Aybars UĞUR, Özgü CAN
9th International Artificial Intelligence and Data Processing Symposium (IDAP), Türkiye, Malatya, Uluslararası, Sözlü Sunum, Tam Metin Bildiri, 2025-2025, Scopus, ss 1-6
DOI: 10.1109/IDAP68205.2025.11222381
Özet
<p><span style="color: rgb(51, 51, 51); background-color: rgb(255, 255, 255);">Attention Deficit Hyperactivity Disorder (ADHD) is one of the most prevalent neurodevelopmental disorders in children and adolescents, characterized by symptoms of inattention, hyperactivity, and impulsivity. Despite the availability of standardized diagnostic criteria, the current diagnostic process heavily relies on subjective behavioral assessments. Recent advances in neuroimaging and machine learning have opened new avenues for the development of data driven ADHD diagnostic tools. A two-stage analysis was conducted using the ADHD-200 dataset. Preliminary experiments were first performed on the NYU subset, followed by comprehensive evaluations on the full dataset. Throughout this process, model accuracy and generalization were significantly improved through techniques such as class weighting, data augmentation, and hyperparameter optimization. Consequently, the optimized 3D CNN model achieved a test accuracy of 78.7 %, demonstrating promising results compared to previous multimodal imaging studies. These results demonstrate that high classification performance can be attained using sMRI data alone and highlight the feasibility of deep learning-based diagnostic systems, even under constraints of limited data and single-modality imaging. Therefore, further research into AI-assisted diagnosis of ADHD and other psychiatric disorders is both necessary and valuable.</span></p>