v1.2.4-rc1

What's Changed

• GBM postop 1p-5p support, added BrainGrid features by @dbouget in #77
• Backend requirements update and support for the cavity class by @dbouget in #78

What's to come

• More complete post-operative package, including edema and cavity classes, for GBM and LGG
• Better handling of model selection (when multiple inputs and/or multiple timestamps)
• New models (after BraTS 2024)

Full Changelog: v1.2.3...v1.2.4-rc1

v1.2.3

What's Changed

• Moving the models list to Github from Google Drive by @dbouget in #71
• CI update by @dbouget in #72
• Fix for models download by @dbouget in #73
• Added quick-fix for loading radiological volumes with intensity range limited to [0, 255].
• Deprecated Python 3.7, building on Python 3.8

Full Changelog: v1.2.2...v1.2.3

v1.2.2

What's Changed

• Support for multiclass segmentation following the BraTS challenge (i.e., necrosis, contrast-enhancing tumor, and edema).
  For now, only in-house models trained on the 2023 BraTS challenge are supported (available for gliomas, meningiomas, and metastases), expecting four MR scans as input (T1w, T1w-ce, FLAIR, and T2).
• Software settings panel adjustment to switch between single-class or multiclass output (Settings > Preferences > Processing - Segmentation > Output classes)

Changelog

• Added simple GUI launch test to CIs by @andreped in #57
• Added GUI launch test in the CIs by @dbouget in #58
• Opencv removal from rt-utils backend by @dbouget in #60
• Minor README refactoring by @andreped in #61
• Brain tumor multiclass segmentation by @dbouget in #62
• Updating build version to Python3.8 by @dbouget in #63
• GUI unit test on Unix systems by @dbouget in #64

Full Changelog: v1.2.1...v1.2.2

v1.2.1

What's Changed

• Support for saving segmentation results as DICOM RTStruct.
• Included post-segmentation refinement options (e.g., inducing over-segmentation by a set margin).
• Improvement to the Settings panel.

Changelog

• Added feature request issue template by @andreped in #37
• Major README refactoring + citation update by @andreped in #46
• Fix: Only bundle one python version; upgraded PyInstaller version by @andreped in #47
• Fixed PySide6 bug during program launch; upgraded PySide6 version by @andreped in #48
• Renamed ARM CI; renamed macOS packages; added ARM CI badge to README by @andreped in #50
• Upgrade opencv-python==4.5.5.64 [no ci] by @andreped in #52
• Fixed CI badge hyperlinks by @andreped in #53
• Badge status updates on all CI run events by @andreped in #54
• Improved citation section in README [no ci] by @andreped in #55

Full Changelog: v1.2.0...v1.2.1

v1.2.0

What's Changed

1. Logic

   • Inference backend swapped from TensorFlow to ONNX Runtime.
   • All segmentation models have been re-trained using a patch-wise approach (160x160x160 voxels) for boosting recall performance, keeping the same AGU-Net architecture.
   • Ealry post-operative glioblastoma rest tumor segmentation has been included.
   • Surgical standardized reporting included for glioblastoma containing pre-/post-operative volumes and extent of resection assessment.

2. GUI

   • Overall palette updated, better consistency for widgets design overall, use of icons from Figma, new software logo.
   • Possibility to load patient data corresponding to multiple timestamps (e.g., pre-operative and post-operative).
   • Improvement for the study/batch mode to hold and display all patients' statistics, and export them to csv.
   • Redesign of the single patient mode to allow for multiple timestamps, and a dedicated actions tab.
   • User preferences/settings widget storing convenience choices regarding segmentation and reporting.

3. Deployment

   • Support for Apple M1 Chip computers.
   • C++ ANTs backend usage for Linux and Mac versions, which in theory provides similar registration results as the Python-wrapped version (used for the Windows version).
   • All trained models are now hosted on GitHub rather Google Drive.

• Refactored build setup by @andreped in #32
• updated ICO and ICNS logos by @andreped in #35
• Remove .raidionics/ directory during install and uninstall for Windows by @andreped in #36

Full Changelog: v1.1.0...v1.2.0

v1.2.0-beta

What's Changed

1. Logic

   • Inference backend swapped from TensorFlow to ONNX Runtime.
   • All segmentation models have been re-trained using a patch-wise approach for boosting recall performance, keeping the same AGU-Net architecture.
   • Postoperative rest tumor segmentation has been included together with postoperative clinical reporting including extent of resection assessment.

2. GUI

   • Overall palette updated, better consistency for widgets design overall, use of icons from Figma, new software logo.
   • Possibility to load patient data corresponding to multiple timestamps (e.g., preoperative and postoperative).
   • Improvement for the study/batch mode to hold and display all patients' statistics, and export them to csv.
   • Redesign of the single patient mode to allow for multiple timestamps, and a dedicated actions tab.
   • User preferences/settings widget storing convenience choices regarding segmentation and reporting.

3. Deployment

   • Support for Apple M1 Chip computers, using the C++ ANTs backend, which in theory provides similar registration results as the Python-wrapped version.

Full Changelog: v1.1.0...v1.2.0-beta

v1.2.0-alpha

What's Changed (planned for the final release)

1. Logic

   • Inference backend swapped from TensorFlow to ONNX Runtime.
   • All segmentation models have been re-trained using a patch-wise approach for boosting recall performance, keeping the same AGU-Net architecture.
   • Postoperative rest tumor segmentation has been included together with postoperative clinical reporting including extent of resection assessment.

2. GUI

   • Overall color palette updated, better consistency for widgets design overall, use of icons from Figma, new software logo.
   • Possibility to load patient data corresponding to multiple timestamps (e.g., preoperative and postoperative).
   • Improvement for the study/batch mode to hold and display all patients' statistics.
   • Redesign of the single patient mode to allow for multiple timestamps, and a dedicated actions tab.
   • User preferences/settings widget storing convenience choices regarding segmentation and reporting.

3. Deployment

   • Support for Apple M1 Chip computers, using the C++ ANTs backend, which in theory provides similar registration results as the Python-wrapped version.

N-B: the latest segmentation models, including the post-operative segmentation model, will only be available in the official release.

Full Changelog: v1.1.0...v1.2.0-alpha

v1.1.0

What's Changed

• Complete overhaul of the GUI for better and more intuitive user-experience.
• Created backend libraries for the core processing.
• Scalable batch mode support.
• Fixed several bugs when installing software across all operating systems.
• Added tutorials in wiki tab.

Full Changelog: v1.0.0...v1.1.0

v1.1.0-alpha

Complete overhaul of the graphical user interface for better and more intuitive user experience.
The backend logic for automatic tumor segmentation, features computation, and standardized reporting have been left untouched.

PS: automatically attached source code is wrong, the current version corresponds to the design_update branch.

v1.0.0

This is the first official release of Raidionics, which is a software for automatic segmentation of brain tumors and computation of brain tumor features from MRIs.

The current version was the one used to conduct the experiments in the study "Preoperative brain tumor imaging: models and software for segmentation and standardized reporting" by Bouget et al. (2022).

The software and trained models are currently implemented for pre-operative MRIs and supports:

• Automatic extraction of brain tumor features presented in a standardized report
• Automatic segmentation of four different brain tumor types (high and low-grade gliomas, metastases, and meningiomas)
• Automatic brain extraction and atlas-based segmentation of other relevant structures
• Visualization of MRI together with (predicted) segmentations in 2D viewers
• Support for batch mode - ability to run analysis on a set of MRIs sequentially without user intervention
• Automatic logging and prompts for presenting analysis steps and final report

The release includes binary installers for Windows (>= v10, 64-bit), macOS (>= Catalina), and Ubuntu Linux (>= 18.04). Simply download the installer specific for your operating system.

To run the software, carefully read through the installation procedure before running (see here).

Full Changelog: https://github.com/dbouget/Raidionics/commits/v1.0.0