Trouble identifying yeast cells

I’m new to CellProfiler, which I’m turning to because Volocity has trouble picking out yeast cells, especially when they are clumped or budding off, or the halo isn’t quite bright enough. I should also start of by saying I need to identify cells in the DIC channel, but I have three channels of fluorescence I also need to analyze too (DAPI, MitoTracker, GFP). But my immediate problem is the DIC channel. So far I’ve just been able to get CP to find bits of the outline of cells, but it hasn’t been picking them out as objects because they are too small. And when I increase the size requirement, it picks out these huge blobs of background that don’t really correspond to the cell locations, and I’ve tried using the automatic methods, so I’m not sure where to go from here. I’ve attached an example image.


Usually DIC images are quite difficult to segment, but yours actually worked out pretty well. In the future, we hope to have a DIC/phase specific module. For now, I used a series of modules to enhance the edges and then the interior of the yeast cells so that they look like white round dots on a black background- which is how CellProfiler looks for cells. The ‘background’ method of thresholding works best on these images, because the noisy background and variable foreground confuse some of the other methods (this is likely what resulted in the one large object you saw).

As far as ‘de-clumping’ the yeast, using the “Shape” and “Distance” methods in combination is best, because this method looks for little creases in the perimeter- much like the neck of budding yeast. To get the de-clumping just right, you may have to play with the “Smoothing Filter” size (I have it set at 20; making it larger will de-clump less, making it smaller will de-clump more, but result in possibly false objects).

hope this helps~
yeast_DIC_pipe.cp (4.08 KB)

This is really helpful, thank you!

With declumping, the Shape/Distance method generally works great, but sometimes it sees three cells clumped together and puts a fourth in between them. Adjusting the smoothing filter size doesn’t seem to have much effect, so I’m not sure how to separate them properly.


Try changing the threshold correction factor to 0.8. There is a lot of variability in the foreground of your image (ie, the cells). Some are pretty bright, others, quite dim. Even though I use Background Adpative- which finds a different threshold for different areas of the image depending on the surrounding background pixels- it’s clear that sometimes neighboring yeast cells are of very different intensities. Lowering the threshold (via the threshold correction factor) helps the algorithm find these dimmer cells and correcting identify them as cells- leading to better de-clumping in the second step.

I didnt have your exact image where you showed the problem, but this seemed to work on the image I had.