TJDR Dataset Overview

Dataset Description

• Name: TJDR: A High‑Quality Diabetic Retinopathy Pixel‑Level Annotation Dataset
• Purpose: To support lesion segmentation for diabetic retinopathy (DR) and enable AI‑assisted DR grading.
• Content: 561 color fundus images captured at high resolution from Tongji Hospital, Tongji University, using various cameras (e.g., Topcon TRC‑50DX, Zeiss CLARUS 500).
• Privacy: Personal information has been carefully removed while preserving clear views of the optic disc, retinal vessels, and macula.
• Annotations: Pixel‑level labels for four common DR lesions—micro‑aneurysms (MA), hemorrhage (HE), hard exudates (EX), and soft exudates (SE)—produced with the Labelme tool by experienced ophthalmologists.
• Quality Assurance: Annotations were performed by senior ophthalmologists to ensure high data quality.
• Data Split: The dataset is divided into training and testing subsets.

Dataset Usage

• Download Link: TJDR Dataset Download

Citation

• Paper: TJDR: A High‑Quality Diabetic Retinopathy Pixel‑Level Annotation Dataset
• Authors: Jingxin Mao, Xiaoyu Ma, Yanlong Bi, Rongqing Zhang
• Published: arXiv preprint arXiv:2312.15389, 2023
• BibTeX:

@article{mao2023tjdr,
  title={TJDR: A High‑Quality Diabetic Retinopathy Pixel‑Level Annotation Dataset},
  author={Mao, Jingxin and Ma, Xiaoyu and Bi, Yanlong and Zhang, Rongqing},
  journal={arXiv preprint arXiv:2312.15389},
  year={2023}
}