Skip to content

Commit 0c21c8b

Browse files
authored
Update Release Notes for version 4.4.0
1 parent 58219cc commit 0c21c8b

File tree

1 file changed

+33
-0
lines changed

1 file changed

+33
-0
lines changed

ReleaseNotes.md

Lines changed: 33 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,38 @@
11
# ITK-SNAP Release Notes - Version 4.2
22

3+
## Version 4.4.0
4+
5+
### New Features
6+
7+
- Integration with nnInteractive ([Isensee et al., 2025](https://arxiv.org/abs/2503.08373)) provides AI-assisted segmentation.
8+
- See [web-based documentation](https://itksnap-dls.readthedocs.io/) and [demo video](https://www.youtube.com/watch?v=gpQ19OPUm34) of this new feature
9+
- Supports point, scribble (paintbrush) and lasso (polygon) modes
10+
- Uses network communication to execute deep learning components on a remote or cloud-based GPU
11+
- Improved support for externally-loaded 3D/4D meshes:
12+
- External meshes are visualized in 2D slice views
13+
- New context menu in the 3D view provides quick access to many mesh visualization options
14+
- Camera controls in 3D view extended with import/export, focusing on mesh vs. image
15+
- New high-performance 2D slice rendering code replaces the slower VTK-based introduced in 4.0:
16+
- Direct OpenGL2 hardware rendering on devices that support it (most modern hardware)
17+
- Fallback to software-based rendering for other devices
18+
- Much faster performance for large images
19+
20+
### Bug Fixes
21+
- Discovered inaccuracies with previous tdigest implementation (fast quantile/histogram estimation) and replaced with code from the Apache project.
22+
- Fixed incorrect histogram computation for mesh data arrays
23+
- Various bug fixes (see GitHub)
24+
25+
### Programmatic Improvements
26+
27+
- New [Deep Learning Segmentation (DLS)](https://itksnap-dls.readthedocs.io/) framework :
28+
- Extends ITK-SNAP with Python-based deep learning capabilities
29+
- DLS communication layers supports network-based connections and SSH tunneling
30+
- Lays the foundation for more scalable integration of additional deep learning inference backends
31+
- Software build automation via GitHub Actions
32+
- Changes to CMake build to use Qt-provided bundling utilities instead of fixup_bundle
33+
- Extensive refactoring to wrap most user-presented strings in `tr()` calls for internationalization
34+
- Translation code uses OpenAI API for automated initial translations
35+
336
## Version 4.2.2
437

538
### Programmatic Improvements

0 commit comments

Comments
 (0)