<b>sMRI-based Diagnosis</b>

The model achieved an accuracy of 98.66%, outperforming existing fMRI- and EEG-based models. This highlights the potential of structural MRI in ADHD diagnosis, reducing dependence on more complex and resource-intensive imaging modalities. For ADHD segmentation, the Improved SwinUNet model was introduced, incorporating the CSIA module and ESDR module. The model achieved state-of-the-art segmentation performance with an IoU of 97.32% and a DSC of 97.50%.