Analyzing ring-shaped vesicles with ParticleSizer

Dear @twagner ,
first of all thank you for this great plugin. I am using it in vescicles sizing and it works very well.
At the moment, I am struggling to analyse these kind of images:

This is the raw image:
Group1_2.tif.zip (120.9 KB)

In few words, I am trying to fine tune parameters and methods to identify those “ring-shaped” vescicles but of course it is not easy for your plugin to identifiy “rings” it tries to split them into smaller objects.

Do you think your plugin can manage that with some set of parameters that you can suggest me? or do you think that I am pretending too much from your plugins?

Thank you,
Emanuele Martini

1 Like

Hi @emartini,

maybe @tinevez’s work-in-progress project at https://github.com/tinevez/CircleSkinner might be interesting for you? I haven’t found the time to use it (yet), but applying a Hough transform on your image sounds like sensible thing to me…

Best,
Stefan

1 Like

Dear Stefan,
I will for sure give a chance to that plugin.
At the moment I am trying/comparing two methods one is to use hough transform plugin https://imagej.net/Hough_Circle_Transform and another one is to implement a Matlab/Fiji plugin to use the imfindcircles function of matlab that seems very powerful for my case and the ParticleSizer plugin.

thank you for the suggestion
have a nice day
Emanuele

This doesn’t make sense to me: according to the documentation, imfindcircles uses a Circular Hough Transform, so it’s the same algorithm also used by the Hough Circle Transform plugin and the CircleSkinner suggested by @stelfrich.

I wonder if there are any specific advantages in the implementation that would justify the additional effort of switching to Matlab in this case.

1 Like

in principle @imagejan I fully agree with you, but the fact is that I’ve discovered the fiji plugins about finding circles after I have started to try to identify them with that matlab function, so let’s say I’ve already fine tuned the parameters (that they are not the same of the fiji plugins :wink: ) in matlab implementation of the Hough Transform and also part of the workflow pipeline at the moment is in the matlab code.

Now maybe the question is why dont you move everything to matlab? :slight_smile:
The reason why I cannot switch everything to matlab, it’s because the ParticleSizer plugin works like a charm for the other vescicles.

Now, that I have got some hough transform plugins for fiji for sure I will try to implement everyrthing in fiji with the help of one of those plugins.

have a nice day,
Emanuele Martini

1 Like

Hi all,

Seeing how circular they are, I gave it a try with the Max Inscribed Circles bit and got some interesting results

run("FeatureJ Laplacian", "compute smoothing=1.0");
setThreshold(-10000, -1.0000);

run("Analyze Particles...", "size=0.20-Infinity circularity=0.20-1.00 show=Masks exclude");
run("Options...", "iterations=1 count=3 do=Close");
run("Fill Holes");
run("Max Inscribed Circles", "minimum=5");
selectWindow("Group1_2.tif");
roiManager("Show All");
roiManager("Show All without labels");
4 Likes

thank you @oburri this is very nice,
great plugin!

Gad you find it useful!

For the logic behind it. The small LoG filter enhances the edges of the vesicles and helps the thresholding.

I usually normalize the image before applying the LoG filter so that I can use a fixed threshold for the LoG filter.

The rest I’ll admit was just trying to get a nice enough mask for the Max Inscribed circles, so I’m sure that part can use a bit more thought :smile_cat:

1 Like

yes that makes sense
thanks a lot :wink:
I will give it a try also with other images that I have (let’s say the treatment condition has not ring shaped vescicles … )
and let you know

A post was split to a new topic: Why ImageJ instead of MATLAB