Can you please help me using a mask on ALL my images without the need to generate the mask for every image? The pipeline often fails in generating the right well mask and then I get no measurements from that image
Flatten Background_IXM.cpproj (458 KB)
The project file does not contain any original images. Can you please post a few (both channels each) so we can test them?
My concern with using the same mask for all wells is that the actual well boundary may shift within the image field from well-to-well. Are you certain this does not happen, and/or are you ok with the variability?
Likely, the segmentation of the well can be improved so that you can continue segmenting the well boundary for each, and that is what I would suggest. For similar screens, I have used the “Background” thresholding method in IdentifyPrimaryObjects, with a very low Threshold Correction Factor (~ 0.4). The benefit of using “Background” is that is that since it uses the mode of the intensity distribution, it tends not to be swayed by any bright foreground objects in your well and selects the most common pixel intensity, typically either the background of the well or outside the well. (Yes, three-class Otsu should too, to some extent, but it may be swayed by the foreground too much I suspect). This is also one of the few times in which I recommend using the Lower and Upper bounds on the threshold, set near the typical Well threshold intensity value, so that your ‘odd’ cases like an image with no well boundary, won’t skew your Well threshold too much.