mCellSeg: A Microscopy Cell Instance Segmentation Dataset
收藏Zenodo2026-06-16 更新2026-05-26 收录
下载链接：
https://zenodo.org/doi/10.5281/zenodo.20174259
下载链接
链接失效反馈官方服务：
资源简介：
mCellSeg is an expert-annotated microscopy image dataset designed for benchmarking cell instance segmentation models. The dataset contains images from two human cell lines acquired using differential interference contrast (DIC) and fluorescence microscopy. It captures diverse cellular morphologies, varying confluency levels, and challenging imaging conditions commonly encountered in real laboratory environments.
Dataset Overview
Total images: 300 high-resolution microscopy images
Annotated subset: 200 images (100 HEK-293T, 100 HUVEC) with paired ground-truth instance segmentation masks
Unlabeled subset: 100 additional images without annotations to support semi-supervised or unsupervised learning
Annotation format: Single-channel 2D TIFF masks, where background is labeled as 0 and each cell instance is assigned a unique integer ID
Key Challenges
High cell density and extensive cell-to-cell contact
Indistinct intercellular boundaries caused by cell adhesion
Weak contrast and blurred cell contours
Tiny and slender cellular structures
Large variation in cell morphology and confluency
Related Publication
@article{alam2026cell,
title={Cell segmentation in microscopy images using a SAM-based U-Net architecture and a novel dataset},
author={Alam, Md Shariful and Jackson, Miriam and Lord, Megan and Meijering, Erik},
journal={Computer Methods and Programs in Biomedicine},
volume={285},
pages={109470},
year={2026},
publisher={Elsevier}
}
Acknowledgements
This work was supported by the UNSW RNA Institute and the ARC Future Fellowship (FT220100092).
Corresponding contact: md_shariful.alam@unsw.edu.au
提供机构：
Zenodo创建时间：
2026-05-15
