license: apache-2.0

SA-Med3D-140K [github]

Dataset Summary

SA-Med3D-140K is a large-scale, multi-modal, multi-anatomical volumetric medical image segmentation dataset. It was created to facilitate the development of general-purpose foundation models for 3D medical image segmentation.
The dataset comprises 21,729 3D medical images and 143,518 corresponding masks.
It was gathered from a combination of 70 public datasets and 8,128 privately licensed annotated cases from 24 hospitals.

Supported Tasks

The primary task supported by this dataset is general-purpose, promptable segmentation of volumetric medical images.

It is designed to train and evaluate models that can segment a wide variety of anatomical structures and lesions across different medical imaging modalities.

Citation

If you use this dataset, please cite the associated paper:

@misc{wang2024sammed3dgeneralpurposesegmentationmodels,
      title={SAM-Med3D: Towards General-purpose Segmentation Models for Volumetric Medical Images},
      author={Haoyu Wang and Sizheng Guo and Jin Ye and Zhongying Deng and Junlong Cheng and Tianbin Li and Jianpin Chen and Yanzhou Su and Ziyan Huang and Yiqing Shen and Bin Fu and Shaoting Zhang and Junjun He and Yu Qiao},
      year={2024},
      eprint={2310.15161},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2310.15161},
}