29  HE切片细胞类型

HE切片的细胞类型鉴定

Published

February 7, 2026

看到一篇2025-12-22日上线的预印版文献(1),报道了Classpose,可用于HE染色全片的细胞类型鉴定。

Classpose的logo。图源:https://github.com/sohmandal/classpose

1. 两种细胞分割/cell segmentations

Instance segmentation … aims to determine whether a pixel belongs to a unique cell (1).

以HE切片为例,直观理解就是选切片上的细胞(核)。

Semantic segmentation … aims to determine to which cell type a pixel belongs to (1).

以HE切片为例,直观理解就是确定每个细胞(核)属于哪种类型(比如中性粒细胞、上皮细胞、淋巴细胞等等)。

2. Classpose的原理?

看不懂文献的方法,好像也没有读懂的动力。

3. Classpose能做什么?

Classpose, as easily trainable analog of the Cellpose-SAM model for semantic segmentation (1).

4. Classpose的优点?

We extensively benchmark Classpose against 3 other state-of-the-art methods (Semantic Celllpose-SAM, CellViT++, and Stardist) across 6 datasets (CoNIC, ConSep, GlySAC, MoNuSAC, NuCLS, and PUMA), showing that Classpose consistently outperforms all other methods (1).

5. Classpose如何使用?

… provide a commond-line tool and a QuPath extension for whole slide-image … (1)

6. 自己安装Classpose的QuPath extension体验下

具体安装过程有些折腾,但最终还是成功了。安装过程参考(2) (3)

Pretrained models有:conic、consep、glysac、monusac、nucls、和puma。

不同pretrained models可预测的细胞类型。

用于测试的HE切片。

Classpose的测试结果(用的pretrained model是conic)。Conic可区分的细胞类型是:neutrophils、epithelial cells、lymphocytes、plasma cells、eosinophils、和connective tissue cells (4)。不同颜色表示不同的细胞类型。

7. 期待Classpose文章正式发表

目前该软件工具还是预印版,期待正式在某个科研杂志发表,这样使用起来更加有底气些。

本着严谨的态度,感觉是不是应该比较下Classpose的预测结果多色免疫荧光染色的结果一致性?

或者基于Classpose分析得到了某种结论,再针对该结论通过某种合适的实验验证下?

给我买杯茶🍵

References

1.
S. Mandal, J. G. de Almeida, N. Papanikolaou, T. A. Graham, Classpose: Foundation model-driven whole slide image-scale cell phenotyping in h&e. doi: 10.64898/2025.12.18.695211 (2025).
2.
3.
W. Kang, josegcpa, Classpose: Foundation model-driven whole slide image-scale cell phenotyping in h&e (with QuPath extension!) (2026). https://forum.image.sc/t/classpose-foundation-model-driven-whole-slide-image-scale-cell-phenotyping-in-h-e-with-qupath-extension/118516/13.
4.
S. Graham, M. Jahanifar, A. Azam, M. Nimir, Y.-W. Tsang, K. Dodd, E. Hero, H. Sahota, A. Tank, K. Benes, N. Wahab, F. Minhas, S. E. A. Raza, H. E. Daly, K. Gopalakrishnan, D. Snead, N. Rajpoot, Lizard: A large-scale dataset for colonic nuclear instance segmentation and classification (2021). https://doi.org/10.48550/ARXIV.2108.11195.