DME Classification in Fundus Imaging with Vision Transformers and ETDRS-Guided Annotation

Published: 15 June 2026| Version 1 | DOI: 10.17632/sdcfn87pvs.1
Contributors:
,
,
,
,
,
,
,
, Juan P. Guerra-Félix, Lenin D. Ochoa de la Paz, Rosario Gulias-Cañizo, Amabile A. Velo-Silvestre

Description

Application of ETDRS-based dataset annotation, in which macular landmarks are defined using a YOLO-based method, to support ViT/BEiT for diabetic macular edema (DME) analysis with a multiclass approach.

Files

Steps to reproduce

To reproduce the results is required to read the paper "Advancing DME Classification in Fundus Imaging with Vision Transformers and ETDRS-Guided Annotation" in Peer-Review Process. "ETDRS landmarks with YOLO" contains the setup to replica experiments and the results to be used to train the Trasnformers Classifier in folder "4-classes model". "4-classes model" contains the setup to replicate the experiments, the results and metrics evaluations. "Ophthalmologist Results" file contains the inter-rater ophthalmologist evaluation.

Institutions

Categories

Edema, Image Classification, Transformer Network

Funders

  • Universidad Autónoma de Baja California
    Grant ID: 105/6/C/14/26

Licence