TN3K
The TN3K (Thyroid Nodule Region Segmentation Dataset) comprises 3493 ultrasound images from 2421 patients, captured between January 2016 and August 2020. These images were selected from over 30,000 images provided by our collaborating hospitals based on the following criteria: (1) Each image contains at least one thyroid nodule area; (2) Images of lymph nodes or those containing large colored areas are excluded; (3) In cases where multiple images were taken of the same area or from the same pati
toujingbu可视化图片
可视化图片 1
可视化图片 2
数据集元信息
| 维度 | 2D |
| 模态 | ultrasound |
| 任务类型 | segmentation |
| 解剖结构 | Thyroid Nodule |
| 解剖区域 | Head and Neck |
| 类别数 | 1 |
| 数据量 | 3493 |
| 文件格式 | JPG |
文件结构
tn3k/
├── label4test.csv
├── label4trainval.csv
├── test-image
│ ├── 0000.jpg
│ ├── 0001.jpg
│ ├── 0002.jpg
│ └── ...
├── test-mask
│ ├── 0000.jpg
│ ├── 0001.jpg
│ ├── 0002.jpg
│ └── ...
├── tn3k-trainval-fold0.json
├── tn3k-trainval-fold1.json
├── tn3k-trainval-fold2.json
├── tn3k-trainval-fold3.json
├── tn3k-trainval-fold4.json
├── tn3k-trainval.json
├── trainval-image
└── trainval-mask
│ ├── 2871.jpg
│ ├── 2872.jpg
│ ├── 2873.jpg
│ └── ...
└── trainval-mask
├── 0000.jpg
├── 0001.jpg
├── 0002.jpg
└── ...图像尺寸统计
| 统计类型 | 间距 (mm) | 尺寸 |
|---|---|---|
| 最小值 | - |
(216, 217) |
| 中位值 | - |
(412, 337) |
| 最大值 | - |
(1463, 771) |
引用
@inproceedings{gong2021multi,
title={Multi-task learning for thyroid nodule segmentation with thyroid region prior},
author={Gong, Haifan and Chen, Guanqi and Wang, Ranran and Xie, Xiang and Mao, Mingzhi and Yu, Yizhou and Chen, Fei and Li, Guanbin},
booktitle={2021 IEEE 18th international symposium on biomedical imaging (ISBI)},
pages={257--261},
year={2021},
organization={IEEE}
}