---
license: cc-by-nc-4.0
language:
- en
base_model:
- YukunZhou/RETFound_dinov2_meh
pipeline_tag: image-classification
tags:
- Indocyanine Green Angiography (ICGA)
- multiclass classification
- RETFound
---
# RETFound-DINOv2 (Impression)
Classification of angiographic impressions on Indocyanine Green Angiography (ICGA) images.

Multilabel problem (labels):
  * unremarkable changes
  * macular neovascularization
  * central serous chorioretinopathy
  * uveitis
  * polypoidal choroidal vasculopathy
  * pachychoroid pigment epitheliopathy
  * choroidal mass
  * Other
       
This model is obtained by finetuning the pretrained model available in HuggingFace at this [link](https://huggingface.co/YukunZhou/RETFound_dinov2_meh).

Trained on: [AngioReport](https://pubmed.ncbi.nlm.nih.gov/40610046/)

## Usage

Check instruction at the following [repository](https://gitlab.idiap.ch/medai/software/paper/fm-overspecialization)

## Model output structure

`[batch_size, num_classes]` (`num_classes=8`)

## Related publication(s)

```
@article{xu_angioreport_2025,
        title = {{AngioReport}: dataset and baseline methods for fundus angiography report generation},
        issn = {0007-1161},
        url = {https://bjo.bmj.com/content/early/2025/07/02/bjo-2024-327006},
        doi = {10.1136/bjo-2024-327006},
        journal = {British Journal of Ophthalmology},
        author = {Xu, Pusheng and Chotcomwongse, Peranut and Zhang, Weiyi and Chen, Xiaolan and Wu, Xinyuan and Chung, Florence H T and Zhang, Xueli and He, Mingguang and Shi, Danli and Ruamviboonsuk, Paisan},
        year = {2025},
}
@misc{he2021maskedautoencodersscalablevision,
        title={Masked Autoencoders Are Scalable Vision Learners},
        author={Kaiming He and Xinlei Chen and Saining Xie and Yanghao Li and Piotr Dollár and Ross Girshick},
        year={2021},
        eprint={2111.06377},
        archivePrefix={arXiv},
        primaryClass={cs.CV},
        url={https://arxiv.org/abs/2111.06377},
}
```

## Author(s)

* Roberto Pulvirenti, Idiap Research Institute