SA-Med2D-20M

# [SA-Med2D-20M](https://arxiv.org/abs/2311.11969)


The largest benchmark dataset for segmentation in the field of medical imaging.

As is well known, the emergence of ImageNet has greatly propelled the development of AI, especially deep learning. It has provided massive data and powerful baseline models for the computer vision community, enabling researchers to achieve breakthroughs in tasks such as natural image classification, segmentation, and detection. However, in the medical image realm, there lack of such a large dataset for developing powerful medical models.

To address the gap in the medical field, we are introducing the largest benchmark dataset for medical image segmentation. This initiative aims to drive the rapid development of AI in healthcare and accelerate the transformation of computational medicine towards a more inclusive direction.

Please visit the [GitHub](https://github.com/OpenGVLab/SAM-Med2D) page and further exploit the dataset!

Due to data privacy and ethical requirements, we currently only provide access to a 16M dataset. We will keep updating and maintaining this database. Please stay tuned for further updates from us.

## 👉 Filesystem Hierarchy
```bash
~/SAM-Med2D-20M
├── images
| ├── mr_00--ACDC--patient001_frame01--x_0006.png
| ├── mr_t1--BraTS2021--BraTS2021_00218--z_0141.png
| ├── ...
| ├── ct_00--CAD_PE--001--x_0125.png
| ├── x_ray--covid_19_ct_cxr--16660_5_1--2d_none.png
|
├── masks
| ├── mr_00--ACDC--patient001_frame01--x_0006--0000_000.png
| ├── mr_t1--BraTS2021--BraTS2021_00218--z_0141--0011_000.png
| ├── ...
| ├── ct_00--CAD_PE--001--x_0125--0000_002.png
| ├── x_ray--covid_19_ct_cxr--16660_5_1--2d_none--0000_001.png
|
├── SAMed2D_v1_class_mapping_id.json
|
├── SAMed2D_v1.json

```
The SA-Med2D-20M dataset is named following the convention below:
```bash
-images
-{modality_sub-modality}--{dataset name}--{ori name}--{dimension_slice}.png

-masks
-{modality_sub-modality}--{dataset name}--{ori name}--{dimension_slice}--{class instance_id}.png
```
Note: "sub-modality" applies only to 3D data, and when "sub-modality" is "00," it indicates either the absence of a sub-modality or an unknown sub-modality type. "dataset name" refers to the specific dataset name that the case is from. "ori name" is the original case name in its dataset. "dimension slice", e.g., "x_100", indicates the dimension along which we split a 3D case as well as the slice ID in this dimension. If we split a 3D case with axis x and the current slice is 100, then the term can be "x_0100". For 2D datasets, the "dimension_slice id" is uniformly set to "2d_none". "class instance_id", unique to masks, encapsulates both category information and instance id, and the detailed information is stored in the "SAMed2D_v1_class_mapping_id.json" file. For instance, if the category "liver" is assigned the ID "0003" and there is only one instance of this category in the case, the "class instance_id" can be denoted as "0003_000". Besides, the category "liver" in the "SAMed2D_v1_class_mapping_id.json" file is formulated as key-value pair with _python-dict_ format: \{"liver": "0003"\}.

The file "SAMed2D_v1_class_mapping_id.json" stores the information for converting class instances. The file "SAMed2D_v1.json" contains the path information for all image and mask pairs.

## 👉 Unzipping split zip files
Windows:

decompress SA-Med2D-16M.zip to automatically extract the other volumes together.

Linux:

1. zip SA-Med2D-16M.zip SA-Med2D-16M.z0* SA-Med2D-16M.z10 -s=0 --out {full}.zip

2. unzip {full}.zip

## 🤝 免责声明
- SA-Med2D-20M是由多个公开的数据集组成，旨在取之于社区，回馈于社区，为研究人员和开发者提供一个用于学术和技术研究的资源。使用本数据集的任何个人或组织（以下统称为“使用者”）需遵守以下免责声明：
1. 数据集来源：本数据集由多个公开的数据集组成，这些数据集的来源已在预印版论文中明确标明。使用者应当遵守原始数据集的相关许可和使用条款。
2. 数据准确性：尽管我们已经努力确保数据集的准确性和完整性，但无法对数据集的准确性作出保证。使用者应自行承担使用数据集可能带来的风险和责任。
3. 责任限制：在任何情况下，数据集的提供者及相关贡献者均不对使用者的任何行为或结果承担责任。
4. 使用约束：使用者在使用本数据集时，应遵守适用的法律法规和伦理规范。使用者不得将本数据集用于非法、侵犯隐私、诽谤、歧视或其他违法或不道德的目的。
5. 知识产权：本数据集的知识产权归原始数据集的相关权利人所有，使用者不得以任何方式侵犯数据集的知识产权。

- 作为非盈利机构，团队倡导和谐友好的开源交流环境，若在开源数据集内发现有侵犯您合法权益的内容，可发送邮件至([e-mail], [e-mail])，邮件中请写明侵权相关事实的详细描述并向我们提供相关的权属证明资料。我们将于3个工作日内启动调查处理机制，并采取必要的措施进行处置(如下架相关数据)。但应确保您投诉的真实性，否则采取措施后所产生的不利后果应由您独立承担。

- 通过下载、复制、访问或使用本数据集，即表示使用者已阅读、理解并同意遵守本免责声明中的所有条款和条件。如果使用者无法接受本免责声明的任何部分，请勿使用本数据集。

## 🤝 Disclaimer
- SA-Med2D-20M is composed of multiple publicly available datasets and aims to provide a resource for academic and technical research to researchers and developers. Any individual or organization (hereinafter referred to as "User") using this dataset must comply with the following disclaimer:
1. Dataset Source: SA-Med2D-20M is composed of multiple publicly available datasets, and the sources of these datasets have been clearly indicated in the preprint paper. Users should adhere to the relevant licenses and terms of use of the original datasets.
2. Data Accuracy: While efforts have been made to ensure the accuracy and completeness of the dataset, no guarantee can be given regarding its accuracy. Users assume all risks and liabilities associated with the use of the dataset.
3. Limitation of Liability: Under no circumstances shall the dataset providers or contributors be held liable for any actions or outcomes of the Users.
4. Usage Constraints: Users must comply with applicable laws, regulations, and ethical norms when using this dataset. The dataset must not be used for illegal, privacy-infringing, defamatory, discriminatory, or other unlawful or unethical purposes.
5. Intellectual Property: The intellectual property rights of this dataset belong to the relevant rights holders of the original datasets. Users must not infringe upon the intellectual property rights of the dataset in any way.

- As a non-profit organization, we advocate for a harmonious and friendly open-source communication environment. If any content in the open dataset is found to infringe upon your legitimate rights and interests, you can send an email to ([e-mail], [e-mail]) with a detailed description of the infringement and provide relevant ownership proof materials. We will initiate an investigation and handling mechanism within three working days and take necessary measures (such as removing relevant data) if warranted. However, the authenticity of your complaint must be ensured, as any adverse consequences resulting from the measures taken shall be borne solely by you.

- By downloading, copying, accessing, or using this dataset, the User indicates that they have read, understood, and agreed to comply with all the terms and conditions of this disclaimer. If the User cannot accept any part of this disclaimer, please refrain from using this dataset.

## 🤝 Acknowledgement
- We thank all medical workers and dataset owners for making public datasets available to the community. If you find that your dataset is included in our SA-Med2D-20M but you do not want us to do so, please contact us to remove it.

## 👋 Hiring & Global Collaboration
- **Hiring:** We are hiring researchers, engineers, and interns in General Vision Group, Shanghai AI Lab. If you are interested in Medical Foundation Models and General Medical AI, including designing benchmark datasets, general models, evaluation systems, and efficient tools, please contact us.
- **Global Collaboration:** We're on a mission to redefine medical research, aiming for a more universally adaptable model. Our passionate team is delving into foundational healthcare models, promoting the development of the medical community. Collaborate with us to increase competitiveness, reduce risk, and expand markets.
- **Contact:** Junjun He([e-mail]), Jin Ye([e-mail]), and Tianbin Li ([e-mail]).

## 👉 Typos of paper
1. Formula (1) is incorrect, after correction: <img src="https://i.postimg.cc/sXRK4MKh/20231123001020.png" alt="alt text" width="202" height="50">

## Reference
```
@misc{ye2023samed2d20m,
title={SA-Med2D-20M Dataset: Segment Anything in 2D Medical Imaging with 20 Million masks},
author={Jin Ye and Junlong Cheng and Jianpin Chen and Zhongying Deng and Tianbin Li and Haoyu Wang and Yanzhou Su and Ziyan Huang and Jilong Chen and Lei Jiang and Hui Sun and Min Zhu and Shaoting Zhang and Junjun He and Yu Qiao},
year={2023},
eprint={2311.11969},
archivePrefix={arXiv},
primaryClass={eess.IV}
}

@misc{cheng2023sammed2d,
title={SAM-Med2D},
author={Junlong Cheng and Jin Ye and Zhongying Deng and Jianpin Chen and Tianbin Li and Haoyu Wang and Yanzhou Su and
Ziyan Huang and Jilong Chen and Lei Jiangand Hui Sun and Junjun He and Shaoting Zhang and Min Zhu and Yu Qiao},
year={2023},
eprint={2308.16184},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```

# [SA-Med2D-20M](https://arxiv.org/abs/2311.11969)


医学影像分割领域规模最大的基准数据集。

众所周知，ImageNet（ImageNet）的问世极大推动了人工智能，尤其是深度学习的发展。它为计算机视觉社区提供了海量数据与高性能基准模型，使得研究者得以在自然图像分类、分割、检测等任务中实现突破性进展。然而在医学影像领域，尚缺乏此类可用于开发高性能医学模型的大规模数据集。

为填补医学领域的这一空白，我们推出了当前规模最大的医学影像分割基准数据集。本项目旨在推动人工智能在医疗健康领域的快速发展，加速计算医学向更具普惠性的方向转型。

请访问[GitHub](https://github.com/OpenGVLab/SAM-Med2D)页面以进一步探索本数据集！

受数据隐私与伦理规范限制，目前我们仅开放1600万样本数据集的访问权限。本数据库将持续更新与维护，敬请关注后续动态。

## 👉 文件系统层级结构
bash
~/SAM-Med2D-20M
├── images
| ├── mr_00--ACDC--patient001_frame01--x_0006.png
| ├── mr_t1--BraTS2021--BraTS2021_00218--z_0141.png
| ├── ...
| ├── ct_00--CAD_PE--001--x_0125.png
| ├── x_ray--covid_19_ct_cxr--16660_5_1--2d_none.png
|
├── masks
| ├── mr_00--ACDC--patient001_frame01--x_0006--0000_000.png
| ├── mr_t1--BraTS2021--BraTS2021_00218--z_0141--0011_000.png
| ├── ...
| ├── ct_00--CAD_PE--001--x_0125--0000_002.png
| ├── x_ray--covid_19_ct_cxr--16660_5_1--2d_none--0000_001.png
|
├── SAMed2D_v1_class_mapping_id.json
|
├── SAMed2D_v1.json

SA-Med2D-20M数据集的命名遵循以下规则：
bash
-images
-{modality_sub-modality}--{dataset name}--{ori name}--{dimension_slice}.png

-masks
-{modality_sub-modality}--{dataset name}--{ori name}--{dimension_slice}--{class instance_id}.png

注：“子模态(sub-modality)”仅适用于三维数据，当“sub-modality”为“00”时，表示不存在子模态或子模态类型未知。“数据集名称”指该样本所属的具体数据集名称。“原始名称(ori name)”为该样本在其原始数据集中的命名。“维度切片(dimension_slice)”例如“x_100”，表示我们拆分三维样本所沿的维度，以及该维度下的切片ID：若沿x轴拆分三维样本且当前切片为100，则可记为“x_0100”。对于二维数据集，“维度切片ID”统一设为“2d_none”。“类别实例ID(class instance_id)”仅针对掩码文件，同时包含类别信息与实例ID，详细信息存储于“SAMed2D_v1_class_mapping_id.json”文件中。例如，若类别“肝脏”被分配ID“0003”，且该类别在该样本中仅有一个实例，则“class instance_id”可记为“0003_000”。此外，“SAMed2D_v1_class_mapping_id.json”文件中，类别“肝脏”以Python字典格式的键值对形式存储：{"liver": "0003"}。

文件“SAMed2D_v1_class_mapping_id.json”用于存储类别实例的转换信息。文件“SAMed2D_v1.json”包含所有图像与掩码对的路径信息。

## 👉 分卷压缩包解压
Windows系统：
将SA-Med2D-16M.zip解压，系统将自动一并提取其余分卷。

Linux系统：
1. 执行命令 zip SA-Med2D-16M.zip SA-Med2D-16M.z0* SA-Med2D-16M.z10 -s=0 --out {full}.zip
2. 执行命令 unzip {full}.zip

## 🤝 免责声明
- SA-Med2D-20M由多个公开数据集组成，旨在取之于社区、回馈于社区，为研究者与开发者提供用于学术与技术研究的资源。使用本数据集的任何个人或组织（以下统称“使用者”）需遵守以下免责声明：
1. 数据集来源：本数据集由多个公开数据集组成，相关数据集的来源已在预印本论文中明确标注。使用者应遵守原始数据集的相关许可协议与使用条款。
2. 数据准确性：尽管我们已尽力确保数据集的准确性与完整性，但无法对其准确性作出任何保证。使用者需自行承担使用数据集所带来的全部风险与责任。
3. 责任限制：在任何情况下，数据集提供者及相关贡献者均不对使用者的任何行为或由此产生的结果承担责任。
4. 使用约束：使用者使用本数据集时，需遵守适用的法律法规与伦理规范。不得将本数据集用于非法、侵犯隐私、诽谤、歧视或其他违法或不道德的用途。
5. 知识产权：本数据集的知识产权归原始数据集的相关权利人所有，使用者不得以任何方式侵犯数据集的知识产权。

- 作为非营利机构，我们倡导和谐友好的开源交流环境。若您发现开源数据集中存在侵犯您合法权益的内容，可发送邮件至([e-mail], [e-mail])，邮件中需详细描述侵权相关事实，并提供相关权属证明材料。我们将在3个工作日内启动调查处理机制，并采取必要措施（如下架相关数据）进行处置。但请确保您的投诉真实有效，否则由此采取措施产生的不利后果将由您自行承担。

- 通过下载、复制、访问或使用本数据集，即表示使用者已阅读、理解并同意遵守本免责声明的全部条款与条件。若使用者无法接受本免责声明的任何部分，请勿使用本数据集。

## 🤝 致谢
- 我们感谢所有医务工作者与数据集所有者将数据集公开分享给社区。若您发现您的数据集被纳入SA-Med2D-20M而不希望如此，请联系我们移除相关内容。

## 👋 招聘与全球合作
- **招聘**：我们正在上海人工智能实验室通用视觉组招聘研究员、工程师与实习生。若您对医疗基础模型与通用医学AI（包括基准数据集设计、通用模型构建、评估系统开发与高效工具研发）感兴趣，请联系我们。
- **全球合作**：我们致力于重新定义医学研究，打造更具普适性的医疗模型。我们的团队正深耕医疗基础模型，推动医疗社区的发展。与我们合作可提升竞争力、降低风险并拓展市场。
- **联系方式**：何俊军([e-mail])、叶锦([e-mail])、李天斌([e-mail])。

## 👉 论文勘误
1. 公式(1)存在错误，修正后如下：<img src="https://i.postimg.cc/sXRK4MKh/20231123001020.png" alt="alt text" width="202" height="50">

## 参考文献

@misc{ye2023samed2d20m,
title={SA-Med2D-20M 数据集：搭载2000万掩码的二维医学影像分割一切任务},
author={Jin Ye and Junlong Cheng and Jianpin Chen and Zhongying Deng and Tianbin Li and Haoyu Wang and Yanzhou Su and Ziyan Huang and Jilong Chen and Lei Jiang and Hui Sun and Min Zhu and Shaoting Zhang and Junjun He and Yu Qiao},
year={2023},
eprint={2311.11969},
archivePrefix={arXiv},
primaryClass={eess.IV}
}

@misc{cheng2023sammed2d,
title={SAM-Med2D},
author={Junlong Cheng and Jin Ye and Zhongying Deng and Jianpin Chen and Tianbin Li and Haoyu Wang and Yanzhou Su and Ziyan Huang and Jilong Chen and Lei Jiang and Hui Sun and Junjun He and Shaoting Zhang and Min Zhu and Yu Qiao},
year={2023},
eprint={2308.16184},
archivePrefix={arXiv},
primaryClass={cs.CV}
}

搜集汇总

背景与挑战

背景概述

SA-Med2D-20M是医学影像分割领域最大的基准数据集，旨在推动医疗AI的发展，目前提供16M的数据访问。数据集包含图像、掩码文件及相关的映射和路径信息文件，组织结构清晰。