Multi-Scale Time-Frequency Fusion Intelligent Diagnosis Model for Children with ADHD Based on fNIRS

Early diagnosis of attention deficit hyperactivity disorder (ADHD) in children and understanding its underlying mechanisms have become key research areas. This study presents a multi-scale time-frequency fusion intelligent diagnosis model (MsTFFM) based on functional Near-Infrared Spectroscopy to enhance diagnostic accuracy for children with ADHD. Data from 40 children with ADHD and 42 healthy controls were analyzed. t-SNE visualization was employed to illustrate diagnostic results, and a “leave-one-ROI-out” method was used to investigate functional abnormalities in specific brain regions. Results demonstrated that MsTFFM significantly outperformed comparison models $(\mathrm{T} > 0,\ p <. 042)$. Additionally, functional abnormalities in homotopic and right hemisphere brain may be linked to the neurophysiological mechanisms of ADHD. This research offers new insights and methodologies for the clinical diagnosis and exploration of ADHD.