search

Neural-Systems-at-UIO / nutil

https://nutil.readthedocs.io/en/latest/
GNU General Public License v3.0
12 stars 5 forks source link

Wrong rat brain area labelling after running the QUINT-workflow with the WHSRat-v3-atlas #214

Closed Wissewald closed 3 years ago

Wissewald commented 3 years ago

I was trying to test run the QUINT-workflow with Ilastik-segmentation of a confocal image and a brightfield image (of the same slice) registered in the WHSRat-v3-atlas. The problem is that the brain areas are wrongly labelled; e. g. "corpus callosum and associated subcortical white matter" is labelled as "olfactory bulb", "neocortex" as "pyramidal decussation" and "hypothalamic region" as "brainstem".

I would be gratefull for any advice. Thanks in advance!

Wissewald

MATLAB_plot+QuickNII_segmentation

leuat commented 3 years ago

Are you certain that you've used the correct WHSRat atlas version (2015 etc) when a) using quicknii and b) using nutil? If not, then labels will typically be completely incorrect, as the ID values are different for the different atlases

Wissewald commented 3 years ago

Yes, I'm sure I used the same atlas in QuickNII and Nutil. What may have caused the problem is that I used an overlay of 2 Ilastik-segmentations and specified the object color as the color of the overlay. But rerunning Nutil with an 8-bit-png-file of a simple segmentation in Ilastik (no overlay, just black objects on white background) resulted in the same wrong labelling.

Wissewald commented 3 years ago

I tried non-linear registration in VisuAlign with the xml-file from QuickNII and then used the resulting json-file for Nutil Quantifier. It worked, as you can see, the areas are labelled correctly and also the colors match with the WHSRat-v3-atlas.

untitled

Majpuc commented 3 years ago

Hi, This is very helpful! thanks! We might have a bug in the QuickNII version of WHSv3 then. We will look into this!

sharoncy commented 3 years ago

@Wissewald Would you be able to send an atlas map, a matching segmentation, the XML and the Nut file for the first erroneous analysis to s.c.yates@medisin.uio.no please? I'd like to investigate.

Did the second analysis you did resolve your issue, or is it important for you to get it to work without VisuAlign? I will investigate either way, just wondering where you are in terms of your analysis.

Thanks!

sharoncy commented 3 years ago

Wissewald sent me some files and I tested. I didn't manage to duplicate the error, the analysis worked as it should including the coordinate extraction, so I don't think there is a bug here. I did update some settings in his NUT file. He is happy with this now so I will close this issue. Sharon Yates 28.07.21.

sharoncy commented 3 years ago

Validated 25.08.21.