Enhancing accuracy and explainability in colorectal lesion classification with attention-supervised Vision Transformers
Direct supervision of ViT attention with L<sub>LFA</sub> leverages expert knowledge to jointly boost Paris classification accuracy and spatial interpretability, and compares favourably with Grad-CAM-based explanation regularisation. The source code and dataset splits are publicly available at https://github.com/LucaCarlini/SUNDatasetPretraining.
Authors
- Davide Massimi (ORCID: https://orcid.org/0000-0002-0926-514X)
- Chiara Lena (ORCID: https://orcid.org/0009-0003-0326-0342)
- Tommy Rizkala
- Luca Carlini
- Cesare Hassan
- Luca Di Stefano
- Elena De Momi
Institutions
- IRCCS Humanitas Research Hospital (IT)
- Politecnico di Milano (IT)
Publication Details
- Journal
- Computer Methods and Programs in Biomedicine
- Published
- 2026-01-21
- DOI
- https://doi.org/10.1016/j.cmpb.2026.109260
- Primary Topic
- Colorectal Cancer Screening and Detection
- Type
- article
- Field-Weighted Citation Impact
- 0.00