Improving cell/nuclie detection in cell counting pipeline


Dear CP users,

I have been trying to optimize my pipeline to count cells dyed with CC1 which is a cytoplasmic dye in sub regions from the images (I use mask to specify where to count). The images I need to count are of different qualities unfortunately, thus I have been tweaking the pipeline to improve images with noisy background. In some of the images however, the nuclei is segmented and DAPI staining isn’t strong, so the pipeline is either removing them in the noise-removal module (or filtered during id primary) or they become false multiple objects. Is there a way to make the pipeline detect those as one object. As well, I can’t get my pipeline to detect all the cells or it detect one cell as many.
I have uploaded the pipeline, one image example with its mask and the result overlay.
I really appreciate any input.

Counting_ROI.cpproj (559.3 KB)


A link to 3 sample images.


I would start the pipeline with RescaleIntensity which standardizes the input and should help keep results consistent from sample to sample. In IdentifyPrimaryObjects I would use Otsu thresholding method. To help equalize intensities in order to include all cells if they have different intensity for further calculations I would use CorrectIlluminationCalculate and CorrectIlluminationApply.