MAD-Former: A Traceable Interpretability Model for Alzheimer’s Disease Recognition based on Multi-patch Attention
We will make the Mad-Former model and interpretability module code publicly available upon acceptance of the manuscript.
The supplementary material provide the significant brain tissues in the 3D ROI with a patch size of 8, as well as the subject IDs used in the manuscript experiments.
2024.1.22 Supplemental generalization evaluation and schematic diagram of the 3D BFEN structure.
We provide pre-trained weights and model prediction modules.
Please contact [email protected] for the pre-trained weights!