Vacuole measurement

Dear ImageJ forum,

My team and I would like to measure vacuole count and/or their diameters across different time frames. The vacuoles in the image above are the black round circles inside a cell (e.g. the ~5-6 vacuoles that form a “ghost face” in the top left section of the image. I am relatively new to Fiji, and I tried using the Trackmate plugin, but it couldn’t detect the vacuoles, I tried using the manual tracking but it is too slow and tedious. We also tried counting them manually but this also could become a large source of error from user to user. Is there something else that you could recommend us to use to make these measurements? Any help is highly appreciated! Thank you !

It is a tricky problem, I would need to try some things to see what can work.
But the high variability in the image and that your objects are not very clearly distinct from the background would create issues for sure.

Maybe could you provide us with a proper image or a crop of some vacuoles with background?
Then I can test some things.

Thank you Christopher for replying to my message!!! I am attaching the original image and a crop of it showing some vacuoles with background. Please let me know if you need anything else! I am looking forward to your image.tif (2.0 MB) Crop_forum image.tif (2.9 MB)


I am trying a couple of things at the moment. The background is a bit tricky for a robust detection.
Could you specify a bit more the objects you recognize as relevant?

You outlined the larger objects here as important:

But what about smaller objects. For instance the smaller dark spots in the upper left corner of the big field of view:

Also what kind of measurements are key for you or what do you want to perform afterwards?
You mentioned count, diameter but also tracking? Is there a time series then and you need to use that for tracking?

Yes the smaller black blobs would also be of interest. For now, we would be interested in just counting them and perhaps measuring their diameter if possible. Tracking is not that important at the moment. Could it be possible to somehow use a combination of find minima, filter to get the background and filter to find circularity? I think the circularity is the main difference between background and the blobs.
Thank you Chris.

I tried the usual stuff for detection with Gaussian Blur and Maxima Detection and then also Laplacian of Gaussian (LoG3D) with a maximum detection. Non of this really gave me good results. These would also then mean you have no actual segmentations. So there would be no direct way to get the diameter, or do any filtering. You could then do some gaussian fit around the detection to get a diameter.

A intensity based segmentation is not really good in this case anyways.

With such a complex background I would go for a machine learning approach. Not so much to detect the objects but to rather distinguish them from the background. An initial try worked ok I would say:

The upside here is that you can then extract the segmentation and filter the objects using shape and size. Have a look here:

There is also tons of tutorials out there:ân

The weka segmentation is part of Fiji:
Plugins > Segmentation > Trainable Weka Segmentation

ONE TIPP: the edges of your images look very different from the rest. I would exclude this outright from the analysis.

Let me know if you have any questions.

Thank you Chris, I will follow your recommendations and I let you know how it goes.