CelloType: pretrained model weights and example datasets

Minxing Pang; Tarun Kanti Roy; Xiaodong Wu; Kai Tan

doi:10.5281/zenodo.22736356

This archive provides pretrained model weights and processed example datasets for CelloType, a Python framework for cell/nucleus segmentation and cell type classification in tissue images. These resources support the inference and training examples distributed with the CelloType GitHub repository and documentation. ContentsThe archive contains five checkpoint files and three dataset ZIP archives (19.14 GB total). The TissueNet checkpoint supports the fluorescent-image segmentation example; the CRC checkpoint supports the colorectal cancer CODEX segmentation and classification example; and the Xenium checkpoint supports the spatial transcriptomics segmentation example. The MaskDINO Swin-L checkpoint is an upstream COCO-pretrained initialization resource referenced by the training tutorials. An additional checkpoint is supplied as cellpose_model_0001999.pth. Data sourcesThe processed example datasets are derived from TissueNet, the colorectal cancer CODEX dataset from the Garry P. Nolan laboratory, and the 10x Genomics Xenium human lung cancer dataset. The CelloType documentation describes these processed datasets as demonstration resources. Please consult the original datasets for their documentation and reuse terms. UsagePlace the .pth checkpoints in the repository's models/ directory. Place the ZIP archives in data/ and extract the relevant archive. Follow the matching CelloType tutorial for the configuration, input format, and execution instructions. Files are provided separately so users can download only the resources needed for a particular example. SHA-256 checksums are provided in SHA256SUMS.txt. Code: https://github.com/maxpmx/CelloTypeDocumentation: https://cellotype.readthedocs.io/en/latest/Inspected code/documentation revision: eb7dc8417a5c130c31b66eba71c3c3f112c22c7d (this is not a claim about the training-time commit). Related publicationPang, M., Roy, T. K., Wu, X. and Tan, K. CelloType: a unified model for segmentation and classification of tissue images. Nature Methods 22, 348–357 (2025), published online November 22, 2024. https://doi.org/10.1038/s41592-024-02513-1 Original resourcesCRC CODEX: https://doi.org/10.7937/tcia.2020.fqn0-0326TissueNet: https://datasets.deepcell.org/Xenium: https://www.10xgenomics.com/datasets/preview-data-ffpe-human-lung-cancer-with-xenium-multimodal-cell-segmentation-1-standardMaskDINO: https://github.com/IDEA-Research/MaskDINO

CelloType: pretrained model weights and example datasets

Files and links (1)

Abstract

Details

Metrics