Releases: CERN/TIGRE
TIGRE 3.0
TIGRE 3.0
Its been a while since a major release, and the community and me have introduced a variety of important features to TIGRE since version 2.0.
I want to thank to all contributors of code, discussion starters and question askers for helping out to get the software here. TIGRE is still in development, please do join and contribute your code, the community will certainly appreciate it.
Ander Biguri
New Features since v2.6:
- Dual Energy CT tools for monoenergetic projection creation [1]
- Pytorch autodiferentiable engine compatible operators
- Curved detector support (in pre-processing, not operators)
- Improvements in installations
- Diondo helical scan support
- Many, many bug fixes
TIGRE 2.6
TIGRE 2.5
- Add flag to compilation that can disable pinned memory
- Detector offset bug fixes (no circle artifact anymore).
- Fixed latest numpy support.
- CUDA 12 support
- Bug fix in Nikon Loader angles
- Better geometry visualizer in python
TIGRE 2.4
Features
Mayor
- Krylov subspace algorithms. Fast and efficient converging algorithms. [arXiv paper]
- YXLON scanner data loaders
- Detector offset weighting code now in all algorithms
Minor
- Support for newer CUDA
- Algorithms can now return error norm, given a ground truth image
- Important bug fix on large scale recon for python
- Many minor bug fixes
TIGRE 2.3
Features
- Faster filtering for FDK in python
- Better timing info for python algorithms
- Bruker Skyscann Data Loader
- VarianDataLoader (in MATLAB) now has all the preprocesing steps that a Varian scanner performs. "TIGRE-VarianCBCT: Streamlined Open-Source MATLAB-GPU Toolkit for Varian Onboard Cone-Beam CT in Image Guided Radiotherapy" In review (Authors hidden due to review in progress)
- plot_angles() added to python
- Better CUDA support (versions)
- Proton CT to Standard CT code conversion. You can now have proton CT data and transform it such that it approximates standard CT. TIGRE can now do pCT. "Extension of the open-source TIGRE toolbox for proton imaging" In review (Authors hidden due to review in progress)
- Tidier Code.
- Add MVS 2022 support.
- Add circular tomosynthesis and linear tomosynthesis geometries in common_geometry.py
- Merged FISTA and FISTA_mod into one for MATLAB
- SART_TV available in python.
BugFix
- Fix L2 and quality measure arrays bug
- FDK detector offset weights missing a factor of 2
- Fix ASD-POCS initnialization bug
- Fix bug in SART algorithms wrongly indexing
- NikonDataLoader.py bug fixes
- Fix 1/2 pixel offset in backprojection
Note: Most of these changes have been added by a set of fantastic contributors to TIGRE. Huge thanks to all of you. - Ander
TIGRE 2.2
Features
- You can now select which GPUs to run the code on, either by name, or GPUid.
- Improved compilation, now there is no need to edit setup.py/xml files for your GPU/CUDA SDK
- Unify source: there is only one folder with all the CUDA code for both python and MATLAB.
- Detector rotation now supported on parallel beam
- Add NikonDataLoader() to python.
- Code quality optional code added, to be able to maintain decent python.
- Demos for python available
- Add 3D Shepp-Logan head phantom to Python.
- Replace noise generator with CUDA implementation.
BugFix
- Fix bug where compute 5.0 was not being compiled in python
- Some arbitrary rotation angles were not being passed to the kernel
- Reduced memory needed in MATLAB NikonDataLoader
- Mayor bug that impedes many iterative algorithms to work in python. The way order_subsets returns arrays changed, now they are not dtype=object, but dtype=float
- ComputeV errored for triplet angle inputs, fixed
- Fixed many python algorithm bugs (lost track of some)
- Removed many unnecesary files
- Fix and refactor checkDevices
- Fix backprojection type variable name and its default value in Python.
TIGRE 2.1: Bug fix edition
Many, many bug fixes, and some features.
Projection types changed variable name in python.
TIGRE v2.0
This new release of TIGRE contains:
- Varian and Nikon scanner data loaders.
- Python 2 and 3 support in both Linux and Windows
- Multi-GPU is now efficiently supported in python
- Better support for 2D tomography
- FDK detector offset support (wang weights)
- FISTA proximal algorithm added (with TV-proximal)
- Dozens of bug fixes and minor improvements
Multi-GPU
Addition to multi-GPU and arbitrarily large tomographic reconstruction.
Addition of FISTA algorithm
Major speed ups
A fair amount of small bugfixes + functioning Python first release
The MATLAB code has gone through a lot of small bug fixes
A new version of the Python code is added. Not final yet, but the entire structure of python-TIGRE has been changed.