MMAFN: Multi-Modal Attention Fusion Network for Attention-Deficit Hyperactivity Disorder Classification

Attention-Deficit Hyperactivity Disorder (ADHD) is a prevalent developmental psychiatric disorder in children, characterized by difficulties in sustaining attention, elevated activity levels, and impulsive behaviors. Accurate diagnosis is crucial but often complicated by symptom overlap with other conditions. While recent advancements in machine learning and deep learning offer innovative diagnostic approaches, challenges remain in effectively applying multimodal data and cross-modal fusion. This study presents a Multi-Modal Attention Fusion Network (MMAFN) aimed at enhancing ADHD diagnosis by integrating functional magnetic resonance imaging (fMRI), structural magnetic resonance imaging (sMRI), and phenotypic data. A unique tri-modal fusion component facilitates effective interaction among these data types. By incorporating a Mamba module within our attention fusion framework, we significantly improve classification accuracy and enhance interpretability for clinical applications. Our model achieved an accuracy of 83.51 %, with various performance metrics improving by 2.22 % to 6.94 % compared to the current best-performing model. Our research contributes to the field by demonstrating the potential of advanced fusion technologies in refining clinical decision-making processes.