search

labsyspharm / mcmicro

Multiple-choice microscopy pipeline
https://mcmicro.org/
MIT License
98 stars 58 forks source link

Cell segmentation is incorrect on Immunohistochemical staining picture #543

Closed egg124 closed 4 months ago

egg124 commented 4 months ago

It appears that the cell segmentation is incorrect(see screenshot from FIJI).I input a Immunohistochemical staining picture and the cell segmentation result looks like something went wrong. 1 2 3

egg124 commented 4 months ago

This is my params.yml file

>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>The MCMICRO params.yml file<<<<<<<<<<<<<<<<<<<<<<<<<<<<

workflow: start-at: segmentation modules: watershed: version: 1.5.4-large options: mcquant: --masks nucleiRing.ome.tif

clarenceyapp commented 4 months ago

@egg124 The input nuclei image needs to resemble a fluorescence image of nuclei such as Hoechst or DAPI as bright pixels on a dark background. In your image, there appears to be regions of cytoplasm as well, which you would need to suppress before feeding into most segmentation tools. Is your input image an RGB image of color dyes? I suggest running color deconvolution in FIJI to separate the nuclei as much as possible from background, resave as 16-bit image, and try rerunning ONLY the segmentation on this new image first. Once you have verified the mask looks better, run quantification on your color-deconvolved image to completion.

egg124 commented 4 months ago

Thanks for your answering, It really helps me a lot!!!