ITK-SNAP release 4.4.0

[Paul Yushkevich]

Dear Users,

We are happy to announce that ITK-SNAP 4.4.0 has been officially released and is available for download. This new major release (graphical overview, release notes) offers

• AI-assisted segmentation (demo, docs) via integration with nnInteractive
• Faster and more reliable rendering
• Support for overlaying surface meshes over images
• Internationalization (Traditional Chinese, Spanish, German)
• Many bug fixes and improvements

As always, please let us know on the Github issue tracker or the mailing list if you encounter any issues with this release.

Thank you!

The ITK-SNAP Team

[Sang Soo Cho]

Dear ITK-SNAP Team,

Thank you for the release of ITK-SNAP 4.4. It is an amazing program, and I use it extensively. I have encountered a few issues that I wanted to bring to your attention:

1. Crash on recent image load – The program crashes when I try to load an image from the Recent Images tab. Opening the same file through File → Open Image… works fine.

2. AI brush size – When using the Brush Tool in AI mode, the brush size is locked at 1 and cannot be changed.

3. Corpus callosum AI segmentation – The results are not as accurate as shown in the instruction video. Are there specific requirements for image resolution or acquisition conditions for the AI segmentation to work properly? I used local GPU.

Thank you very much in advance for your guidance and for your excellent work on this tool.

Best regards,
Sangsoo

[pyush...@gmail.com]

Hi Sangsoo,

 

Thank you so much for your message and your kind words! We will look into the recent image load crash. As far as the brush size, this was a recommendation of Fabian Isensee from the nnInteractive team, as they have trained the model using single pixel width scribbles.

 

I cannot speak to the CC segmentation specifically. nnInteractive has an auto-zoom feature that takes care of scaling very large images automatically. The CC is not a very well defined structure on T1-MRI, as it is just a set of tracts from the overall white matter, and there is no defined intensity boundary between it and other white matter regions. Does the segmentation improve as you make additional interactions? Is there any difference between using the polygon tool and the brush tool?

 

Thanks!
Paul

--
You received this message because you are subscribed to the Google Groups "itksnap-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to itksnap-user...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/itksnap-users/d296c550-e45c-442f-9781-6b8395e8049an%40googlegroups.com.