Unable to identify large cells and filter small impurities well

[ZhangPenglan]

Hi,

Thanks for this super helpful tool, but I have some problems running deepcell mesmer, it would be great if you can help me.

1. When I run the following code to identify cytoplasm, I find that it can identify smaller cells well, such as immune cells in the liver, but it is not good for hepatocytes (GS staining). Do you have any suggestions for this situation?
2. When identifying cytoplasm, how can I automatically filter out cells smaller than a certain size, such as 50μm, to filter out some impurities?

docker run -it --gpus all -v C:\Data\deepcell:/data vanvalenlab/deepcell-applications:latest-gpu mesmer --nuclear-image /data/ROI.tiff --membrane-image /data/ROI.tiff --output-directory /data/masks --output-name mask.tif --compartment whole-cell

Thank you very much for your time and help, and I look forward to your reply.

[ngreenwald]

Hi @ZhangPenglan,
You can try playing around with the resolution of the input images to see if that helps performance on hepatocytes.

For filtering out small cells, you should be able to take the output segmentation masks and run them through any of the downstream image analysis pipelines (such as QuPath, HistoCAT, ark-analysis, Mantis-Viewer, etc) and filer the cells based on a cutoff.