Problem with nuclear feaures

Hi team,
I’m using CP Version 1.0.5811(windows xp crystal) for my research work on gastric adenocarcinoma. I want to measure nuclear size, perimeter, shape, hyperchromaticity and nucleolar size but using Primautomatic module either it measures things other than nuclei or it segments a single nucleus. I have tried to adjust smoothing filter size but it isn’t working probably due to my inexperience. I m attaching images and pipelines, please have a look and help me out. Waiting for your reply.

Eliza Ali.
DefaultOUT__4.mat (373 KB)DefaultOUT__1.mat (398 KB)

Hi Eliza,

After talking over the images with one of the image assay developers, our opinion is much the same: There is not much in the images to distinguish the nuclei from the surrounding tissue.

Inverting the image and using IdentifyPrimAuto is a good start, though, and you may want to change the smoothing and maxima distance settings in IdentifyPrimAuto from the automatic values and manually set them as actual numbers. Increasing the smoothing filter size is and/or increasing the maxima distance will lead to less segmentation (i.e., less clump division); decreasing the smoothing filter size and/or maxima distance will do the opposite. It is usually best to change these separately rather than together so you can get a sense of the effect of each value. Also, you can put a PauseCellProfiler module in the pipeline to halt execution at the module of interest, rather than running it to completion.

Hopefully, others will have some additional thoughts.


Thank you for your help. Unfortunately i m unable to find the settings for smoothing filter size and maxima separation distance in identify Primautomatic. I have tried different methods, it is giving option for threshhold correction factor but not for these two variables. Can you please tell me about that.

Hi Eliza,

The two variables Mark is referring to are after ‘Method to distinguish between clumped objects…’:
Size of smoothing filter, in pixel units (if you are distinguishing between clumped objects). Enter 0 for low resolution images with small objects (~< 5 pixel diameter) to prevent any image smoothing.
default = Automatic
Suppress local maxima within this distance, (a positive integer, in pixel units) (if you are distinguishing between clumped objects)
default = Automatic

The ‘automatic’ value is determined based on the image size and is displayed to you on the IdentifyPrimAutomatic output window. You can see what value CellProfiler is choosing and decide if increasing it (by entering your own input where it says ‘Automatic’) might help. For example, if CellProfiler is choosing 6 and 7 for the two variables, but your nuclei are much bigger- say 20 or 40 pixels in size- and tend to be nonuniform in texture, you may want to increase the smoothing filter and suppression of local maxima to something closer to the size of your nuclei. That way, if the algorithm finds a blotchy nucleus, it won’t think it is several different nuclei. If you make the smoothing filter and suppression of local maxima numbers too large, CellProfiler will start lumping all your nuclei together.