Suggestions to Improve segmentation results


I have images in co-culture setting(2 different classes). I have built a pipeline to identify and separate the cells based on standard intensity features.
I have also analyzed the results in CPA (as CP results showed under segmentation) and I’m getting an accuracy of 0.84 I don’t have annotated images (manually labelled ground truth) as there are over 3000 cells per image so it doesn’t seem feasible.
I would like to improve my results. Any suggestions to improve my pipeline(attached) would be helpful.

Thanks in advance!

ImprovedPipeline_final.cpproj (150.3 KB)

Hi @Swati,

Could you upload an example image?

1 Like

Hi @lmurphy

Thanks for your reply280120-SUPB1-HS5-WF-10X-Cyquant_C09_s1_w1.TIF (8.0 MB) 280120-SUPB1-HS5-WF-10X-Cyquant_B08_s10_w1.TIF (8.0 MB) 280120-SUPB1-HS5-WF-10X-Cyquant_B08_s9_w1.TIF (8.0 MB)
I’ve attached a couple of images for your reference.

Hi Swati,

I tried suttle changes in your IdentifyPrimaryObject module “Fill holes” options & declumping factor. It is better than before, though not sure if this is what you are looking for. PFA screen shot.

Fujifilm Wako Automation (Consultant)
For CellProfiler training or optimised pipeline write to,

Read more on our site.
Yokogawa CV8000 - The Ultimate in Confocal HCS