- Wrong diffusivity value in the
set()
method of theNODDI
model (typo introduced in f070c23)
- Added
cp312
wheels to github actions
- Restict the
numpy
version to<2.0.0
(due todipy
problems) - Parameters in the
set()
method of the models have now default values MacOS arm64
wheels are now compiled with GitHub Actions instead of Cirrus CI
- Link to installation guide in
README.md
- Improved
print()
statements when running fromJupyter notebooks
- Bump
dmri-dicelib
tov1.0.1
- Implement the
set()
method in theVolumeFractions
model
- From multiprocessing to
multithreading
(huge speedup in models fit! 🚀)models
andlut
modules in Cythonfit()
methods in Cythondir_to_lut_idx()
method in Cython
- New
NODDI
maps names:- from
ICVF
toNDI
- from
OD
toODI
- from
FISO
toFWF
- from
- New output images name casing (e.g. from
FIT_NDI.nii.gz
tofit_NDI.nii.gz
) - Default Diffusion Tensor model fit method to compute the primary direction in each voxel from
WLS
toOLS
- Default number of threads used by BLAS libraries from
-1
(all threads) to1
- Parallel threads config name from
parallel_jobs
ton_threads
- Default
study_path='.'
andsubject='.'
incore.Evaluation()
- Default
ndirs=500
incore.generate_kernels()
- Expire the deprecated
ndirs
parameter incore.setup()
- Replace
tqdm
withdicelib.ui.ProgressBar()
- Switched to proprietary license (see
LICENSE
file)
- C++ interfaces for NNLS and LASSO solvers (provided by
spams-cython
) DTI_fit_method
config to choose the Diffusion Tensor model fit method (from DIPY)OLS
(Ordinary Least Squares)WLS
(Weighted Least Squares)
BLAS_n_threads
config to set the number of threads used in the threadpool-backend of common BLAS implementations (e.g. OpenBLAS)doComputeRMSE
config to compute the Root Mean Square Error imagedoSaveModulatedMaps
to compute modulated maps for the NODDI model (Tissue-weighted mean), e.g.fit_NDI_modulated.nii.gz
- DeprecationWarning:
dipy.reconst.shm.real_sym_sh_basis
(with dipy>=1.4.0)
invalid command 'bdist_wheel'
error when install with pip <= v20.0.2- #155 Replaced the NiBabel
get_data()
method withget_fdata()
(raiseExpiredDeprecationError
with NiBabel >= v5.0.0)
- Problem with the
set_solver()
method that raise amultiple values for keyword argument
error - Replaced
np.bool
,np.float
andnp.object
deprecated aliases (raiseAttributeError
with NumPy >= v1.24.0)
- 'b0_min_signal' parameter in 'load_data()' to crop to zero the signal in voxels where the b0 <= b0_min_signal * mean(b0[b0 > 0])
- 'replace_bad_voxels' parameter in 'load_data()' to replace NaN and Inf values in the signal
- Check if DWI file is a 4D image
- Removed unused 'verbose' option in init (now all is controleld by 'set_verbose')
- Implemented the 'synthesis.py' module for kernels generation
- 'datasynth' tool (CAMINO) no more needed
- Import error in 'util.py' on Windows systems
- Loading errors in 'lut.py' on Windows systems
- Problems with joblib and pickle on some systems (issue #136)
- Added 'packaging' to requirements as it is not present by default on some systems
- Removed config option 'parallel_backend' as only 'loky' works in current implementation
- Bug when using 'doSaveCorrectedDWI' (issue #134)
- Replace all 'print' with 'PRINT'
- Function 'amico.set_verbose' to control what is printed during execution
- Required dependency from 'python-spams' to 'spams'
- Required dependency numpy>=1.12,<1.22
- amico.core.setup: removed 'ndirs' parameter, now precomputes all directions
- Shortcut 'amico.core.setup' to 'amico.setup'
- Removed unused hasISO parameter in experimental VolumeFractions model
- Install information are stored (and taken from) amico/info.py
- Possibility to fit different voxels in parallel
- Config options 'parallel_jobs' and 'parallel_backend' for better control of parallel execution
- Possibility to specify d_perp<>0 for the Stick in StickZeppelinBall
- Forcing SPAMS to use only one thread
- Added missing dependencies, e.g. wheel
- Replaced in-house progress bar to tqdm
- Warning message in util.py fsl2scheme and sandi2scheme
- Controls if files exist before opening
- Wrong datatype when saving results
- SANDI model
- Possibility to set one single direction in the LUT resolution
- Function get_params to get the specific parameter values of a model
- Moved the documentation to the Wiki
- Added the parameter 'd_par_zep' in the StickZeppelinBall model
- Modify setup.py and fix spams dependency
- Checks if input mask is 3D.
- Use d_perps to set models instead of ICVFs.
- Canged case of some parameters for consistency
- Functions to colorize the output messages.
This version of AMICO is not compatible with COMMIT v1.2 of below. If you update AMICO to this version, please update COMMIT to version 1.3 or above.
- Changelog file to keep tracking of the AMICO versions.
- New feature that allows to decrease the LUT resolution. Example here.