Using MaskImage and IdentifyPrimObject

cellprofiler

#1

The primary objects I want to identify are in transiently transfected cells in the green channel. First I wanted to know if anyone has suggestions on how to mask the image (using MaskImage) for different ranges of intensity, and then using those masked images in subsequent modules. I got the maskimage to work, however, when I try and use IdentifyPrimObject only in the image created by the selected objects in a specified range, the identification always gives results in objects outside of the masked image. Currently I am just using crop to rectify this problem, and cropping the image based on my range of expression levels that i want.

Next, I have ran into a wall identifying a specific primary object. Like i said I am identifying primary objects in a channel that is transiently transfected. However, the object that I am identifying is not specifically stained. It is actually the brightest spot within a nucleus (the spot is a multiple copy of promoters). However, the cells that have this bright spot, also have a background intensity in the area of the nucleus that is not the spot (the nucleoplasm), but not always in the same intensity ratio of spot to nucleus. So I used a CorrectIllumination module to take away some of the background, then increased the threshhold on the identifyPrimObjects module, however, we often get false positives and false negatives on the spot identification. Any suggestions would be appreciated. Also if you would like I can send a power point illustrating these questions.

Thanks,
Elizabeth


#2

Hi Elizabeth,

When you are identifying objects with IdentifyPrimAutomatic, are you using the masked image, “MaskBlue” for example? I am not sure I understand exactly what you are doing when you say you are selecting objects in a specified range. Are you identifying objects, using FilterByObjectMeasurement to get objects in a certain intensity range, and then trying to mask with those filtered objects?

If it is always brighter then the rest of the nuclei, you should be able to keep increasing the “Threshold Correction” until it only identifies the brightest part of the nucleus. Have you already tried this?

Regards,
Mike


#3

Hi Mike,
yes I am identifying objects using FilterByObjectMeasurement (measurement being mean intensity) and then filtering them in the desired range of intensity. Then I try and mask those that I filtered by MaskImage, so I can apply subsequent modules to only the objects that I filtered. The only way that I have done this so far is to crop based on the filter.

yes, the spots are always brighter than the rest of the nuclei, but not always the same percentage brighter than the rest of the nuclei (ie the ratio of nucleoplasm to spot varies). This causes problems because if the threshold is too low it results in false positives and then sometimes the threshold is too high then it results in false negatives. An additional problem with just increasing the threshold is that if I have a bulk of images all treated with different treatments, then sometimes the spot is a really definite bright spot, and other times it is a bigger, less definite (more varied with intensity) spot. This causes the pipeline to be really good at identifying one type of spot, but is harder to identify a wider range of spots.

Thanks for your help,
Elizabeth