Cell Segmentation with Globally Optimized boundaries (CSGO) is a whole-cell segmentation pipeline tailored specifically to the H&E domain. CSGO brings four notable contributions:
1) a U-Net algorithm to detect cell membranes with various intensities; 2) integration of HD-Yolo nuclei detection with the U-Net membrane detection; 3) superior performance and robustness: CSGO consistently outperforms the state-of-the-art cell segmentation algorithm across a range of cancer types; all while requiring fewer manual annotations; and 4) this web-based software to facilitate CSGO usage.
We compared our method to the existing state-of-the-art method Cellpose. Our method consistently achieves a higher F1 score across four of the five external datasets.
1. Upload
Prepare and upload your image file2. Analyze
Analyze the uploaded image with our online tool3. Result
Download result imagesIf you use CSGO in your publication, please cite the following paper:
Gu, Z., Wang, S., Rong, R., Zhao, Z., Wu, F., Zhou, Q., ... & Xiao, G. (2025). Cell Segmentation With Globally Optimized Boundaries (CSGO): A Deep Learning Pipeline for Whole-Cell Segmentation in Hematoxylin-and-Eosin–Stained Tissues. Laboratory Investigation, 105(2), 102184.