Help with processing chains for TANGO plugin

Image Analysis
, , ,
#1

Hi there,

I was just about to start using the TANGO plugin for segmentation of my image stack data. However, with the template processing chain provided from the TANGO website the plugin misses most of the nuclei in my data and is not very helpful for now.

I was wondering how sensitive TANGO is for out-of-focus signal , since I acquired the images in a way that there is still some diffracted signal at the bottom and top of the stack…?

Also, does anybody know about some tutorials/documentations or even pre-set processing chains available for download that can help getting started with TANGO? Currently, I find it difficult to set up processing chains that make sense and fit to my data without any comparison or idea which combinations make sense…

Thank you and best wishes,
Patrick

#2

Dear Patrick,

I admit TANGO may be difficult to master at beginning, I am sure you know already the official website : http://biophysique.mnhn.fr/tango/about

Unfortunately we do not have time to update it regularly, but we will be happy to help you to set up your processing chains, if you can upload some sample image.

Best

Thomas

#3

Hi Thomas,

thanks for your reply and help! Yes, I already went over and through your website. But as you mentioned correctly, it might be a bit difficult to get started if you are not too familiar with the different plugins and filters and what they do exactly within the processing chain. But I understand that you don’t have the time to update all that regularly!

However, here is a link to the files I would like to process. The first channel always corresponds to DAPI [nucleus], ch2 = A488 [the signal I want to quantify in the end], ch3 = Ch546 [post-translational histone modification, not important], ch4 = A647 [EdU, also not important for this context].

I get that with overlapping nuclei it is of course difficult and one needs manual segmentation. But I already had difficulties segmenting the nuclei that are well shaped and without any overlap. Even the basic processing chain provided by you as a template on the website did not segment all of those nuclei. I was wondering if this is due to the out of focus signal that touches the Z-edges and if this can be overcome?

I appreciate any help and comment on this and the processing chain you/others might come up with! Thanks you very much for your help!

Best,
Patrick

#4

Hi Patrick,

For non touching nuclei you can easily segment them with simple thresholding. Here the results for the first stack

1035×715 153 KB
.
For the touching nuclei you need to do manual segmentation, just draw around the nucleus for some slices, thanks to interpolation you will get the whole nucleus in 3D. Yes the blur in Z will have the nuclei touch, maybe you cn remove these out of focus slices.
Hope this helps.
Best,
Thomas

#5

Hey @ThomasBoudier,

thanks a lot for spending your time on that! Would you mind sharing the processing steps with me so that I can follow the segmentation steps and get familiar with the different algorithms used to get there? That would be great!

Thank you again very much and all the best,

Patrick

#6

Hi Patrick,

The simplest seps are are as follows :

  1. FastFilters3D Median 4x2
  2. SimpleSegmenter AutoThreshold OTSU, min size 10000 if necessary

Hope this helps

Best,

Thomas

#7

I am using fiji imageJ 1.53c
Java 1.8.0_172 (64bits)
I am getting the following error while either testing from processing chain or from the ‘Run’ from Data tab
thanks, joby

segment field:tangoSampleImage.zvi
process field error: tangoSampleImage.zvi java.lang.ClassCastException: java.lang.Float cannot be cast to java.lang.Integer
tango.plugin.segmenter.NucleusEdgeDetector.runNucleus(NucleusEdgeDetector.java:78)
tango.plugin.segmenter.NucleusSegmenterRunner.run(NucleusSegmenterRunner.java:63)
tango.dataStructure.Field.processNucleus(Field.java:273)
tango.gui.FieldManager.processAndCropFields(FieldManager.java:126)
tango.gui.FieldManager.processAndCropFields(FieldManager.java:636)
tango.gui.FieldManager$2.run(FieldManager.java:601)
//

#8

HI @Joby_Joseph ,

You get this error message because your images are 32-bits, I guess the NucleusEdgeDetector plugin cannot handle 32-bits images. You either need to use another plugin for segmentation or convert your images to 16-bits.

Best,

Thomas

#9

I am using the example image ‘tangoSampleImage.zvi’. And it says it is 16 bit. Any idea?
Thanks for responding.
joby

#10

Hi @Joby_Joseph,

Ah you’re right, it seems something has changed that prevents the NucleusEdgeDetector to work. I do not think I will be able to fix it, since TANGO is not really maintained any longer.

However, most of the other plugins should still work, I suggest you to try another plugin for segmentation, such as the one suggested previously in this post.

Best,

Thomas

#11

Thanks Thomas. That helped.
joby

#12

I proceeded further. Nuclie and Nucleoli gets segmented if I process the cell. And once processed it stays put in the objects list. But centromeres is acting wierdly. If they are chosen as object for a particular cell then Run will fill their list in the objects. However there are two things odd about it. 1) They vanish when we select another cell. 2) They do not get detected in measurement is checked.

#13

Replacing jars form here solved this issue!
https://imagejdocu.tudor.lu/doku.php?id=plugin:stacks:3d_ij_suite:start#download