Dividing lines between identified objects


#1

Hi,

we would like to quantify small DAPI-positive spots around the cell nucleus. In order to detect even the very small ones, which are very close to the nucleus, we decided to first identify the nuclei with relatively stringent settings (plus distinguishing clumped objects) and then to crop the original DAPI-image in the shape of these objects. We would then substract the cropped image from the original and use the resulting image to identify the small spots. However, this doesn’t work properly because the pixels along the lines which divide clumped objects are not treated as object pixels and are therefore not contained in the cropped image and ,in turn,tried
are present in the resulting substracted image.

I have tried to get around this problem by expanding the objects, but they don’t expand at the position of the dividing line.

I don’t know whether it would help to unify objects that are touching before the cropping step, but I think the unify module is not available in the recent versions.

I would be very grateful if anyone could help me.

regards

DS


#2

Hi DS,

You are correct in noting that identified objects are separated by a border of one pixel of zeros. I confirmed this myself recently and will discuss this with our group this week to see if this behavior ought to be changed so that there is no gap between objects. For now, I don’t know of a way within CellProfiler to change this. Of course, the best way would be to image at a higher resolution so that one pixel widths are not an issue, but I don’t know if that is feasible for you.

I’m not sure what you mean by “…and ,in turn,tried are present in the resulting substracted image.” What do you mean by “tried”?

In the meantime, I might suggest that a better way to define small spots (without having to crop out a larger object like the nucleus) is to use this method:
LoadImages
Smooth (using "Remove BrightRoundSpeckles, and a filter size just larger than the expected speckle size)
Subtract (Orig minus Smoothed)
IDPrimAuto (on the subtracted image, with size limits around the small speckle size)

Hope this helps
Dave


#3

Sorry,

this is a mistake derived from cutting and pasting of the text.

Thanks a lot for your suggestion.

best wishes

DS