Active contours in 3D?

Hi,

I am a new user of Icy so I am sorry if this is a very simple question:
I am trying to use the active contours plugin to generate a 3D ROI around a 63x image of a nucleus (DAPI) to measure its volume and shape measurements in 3D. It works very well for single stacks, however I can not figure out how to apply it for the whole confocal z-stacks and generate a 3D mesh ROI.

I have tried to convert the 2D ROI that I use as input for Active Contours to 3D and then use this as input but it does not work.

Any solutions?

Thanks a lot in advance!

Hi Emma,

To use the Active Contours in 3D you need to initialize it through a 3D ROI. Can you initialize automatically your contour using Thresholder plugin for instance so it will create a 3D ROI which can be used to initialize AC ? Otherwise you can always create a 2D ROI than your convert to 3D stack but it may not work as nicely than a generated ROI (from the nuclei for instance) :

image

Hope that helps !

– Stephane

1 Like

Hi Stephane,
Thank you very much for you reply. I am now using ‘convert to 3D stack’ on a 2D active contour generated ROI of the stack where the nucleus has the largest area. This works ok (but not always and crashes icy often) but to create a 3D ROI with thresholding would be better as this method is not great. The problem I have with thresholding (thresholding plugin and K-means) is that it does not detect the nucleus as one object, and if I limit the object size to detect large objects (size just below the nucleus) it does not detect any object. Do you maybe have suggestions to improve the tresholding when there is only one very clear nucleus (DAPI) in the image?
Thanks in advance!
Emma

Hi Emma,

Filtering the image (gaussian filter for instance) and post processing on ROI (fill holes) can help definitely :slight_smile:
Can you provide a sample image so i can give you more details here ?

– Stephane

1 Like

Hi Stephane,
Ok thank you, that would be good to try then, I have attached a sample image (1 image of the stack, I couldn’t upload the whole stack)sample image nucleus EvB 1 slice.tif (2.4 MB) .
Emma

1 Like

Hi Emma,

I made a quick test and here’s the result i obtained on your image :

With this small protocol : test.protocol (3.5 KB)

Of course you need to adjust the object size of HK-Means block for 3D data (so roughly you can multiply the min/max object size by the number of slice corresponding to object depth).

Hope that helps.

– Stephane

1 Like

Hi Stephane,

That is great, it indeed works for me now as well.

Thanks a lot!
Emma

1 Like