Artifacts in nuclei counting

I am using modified version of URL pipeline for PositivePercentage cell counting.

Result as in cell count, in excel data output were substantially more then expected cell no. stained for nuceli with DAPI and FITC (Positive, EdU incorporated).
I have tried adjusting the threshold intensity method and parameters like pixels. Still, the results are not satisfactory.

Suggestion would be very helpful.

Attached files are Excel data output, Images, and pipeline used.

DefaultOUT__1_Image.csv (2.64 KB)
Gray.rar (199 KB)
ExamplePercentPositive.cp (10.3 KB)


Since, the software can be extremely useful for my research and planning to continue thus optimizing the pipelines for few other experiments, timelapse microscopy and xenograft experiments with labelled cells.

In continuation of my previous post-
I have tried further to improve the nuclei counting pipeline for EdU incorporation experiment, by following the forums and guide for Cellprofiler 2.0.

In brief, a set of image includes, DAPI BLUE all stained nucleus, 1st image and out of which some FITC GREEN, EdU proliferating cells, 2nd image. During analysis, in case of DAPI images the single nucleus is fragmented and considered as two whereas in FITC, Green for EdU, simply the counting is more, false positive. I have tried tuning low and high threshold bounds, threshold intensity, measurement of objects etc. Could able to rectify some basic errors and also included some new modules in the pipeline which improved the counting, surely. However, I believe still needs further tuning with some parameters or including few other modules might help, unsure though.

Therefore, suggestions or correction would be of a great help.

Trial 7-PercentPositive.cp (11 KB)
DefaultOUT__7_Image.csv (3.03 KB)
Gray.rar (240 KB)

Hi Dinesh,

I’m attaching a pipeline which should point you in the right direction. The primary change is the manual adjustment of the smoothing filter size and maxima suppression distance (which probably still need some adjusting), and the use of the RelateObjects module to filter for EdU-positive cells. Your images seem to be high-contrast, so no pre-processing was needed.

2012_11_06.cp (9.04 KB)