Aptos - Diabetic Retinopathy Grading

Description

This dataset contains retinal fundus images for diabetic retinopathy grading using the Davis Scale. The task involves classifying the severity of diabetic retinopathy from retinal images into multiple grades. 16 reasoning traces were collected for each example in this task by sampling with GPT-4o, available in the responses column. We greatly appreciate and build from the original data source available at https://www.kaggle.com/datasets/mariaherrerot/aptos2019

Data Fields

• question: The classification question asking to grade diabetic retinopathy
• options: Multiple choice options (A, B, C, D, E) representing different severity grades
• answer: The correct answer/grade
• image: Retinal fundus image
• responses: Model reasoning responses (in train split)

Splits

• train: Training data with model responses
• test: Test data for evaluation

Usage

from datasets import load_dataset

dataset = load_dataset("OctoMed/Aptos")

Citation

If you find our work helpful, feel free to give us a cite!

@article{ossowski2025octomed,
  title={OctoMed: Data Recipes for State-of-the-Art Multimodal Medical Reasoning},
  author={Ossowski, Timothy and Zhang, Sheng and Liu, Qianchu and Qin, Guanghui and Tan, Reuben and Naumann, Tristan and Hu, Junjie and Poon, Hoifung},
  journal={arXiv preprint arXiv:2511.23269},
  year={2025}
}