Make a new release with documentation build flow (#165)

* revised documentation of Cartesian Fourier operators * run isort to optimize imports * Base operator docstring * Use f-string and clean code * rename coefs to coeffs * format docstrings * [typo] Use american learned. * add dtype option for gradient. * [test] add pydocstyle to CI * Use FourierOperatorBase class. * use the uses_sense property * single Channel reconstruction can also be used for SENSE. * add shape and n_coils * fix import * last check function removing * add docstring * uses sense check for Gradient operator. * cleanup doc and add ref * cast to builtin int Make the condition pass if `n_coil` is a custom integer type (eg np.int64) * inherit from OperatorBase * remove redundant declaration of op/adj_op method. * pass extra gradient args to parentclass. This allow to setup deeper parameters, such as input_data_writeable. * add auto threshold operator. * fix deprecated dtype (error with numpy 1.24) * fix bugs. * add auto sure example. * rework of the eestimation function. * cleaning. * remove level-shared method. * add thresh range argument. * someone left a bug here. * refactor: externalize the computation of wavelet threshold. * improve docstrings. * use elif blocks. * s/level/scale/g * improve auto-threshold example. * update docstring parameters. * add import in init modules. * Cartesian-> cartesian * coeffs_shape * Fix typo * feat: update the example for auto_threshold. * Final Documentation Pull Request #155 (#158) * documentation reconstructors * Revised the documentation on non-uniform Fourier operators * Clean documentation on gradient operators * Update mri/operators/fourier/cartesian.py Co-authored-by: Pierre-Antoine Comby <[email protected]> * Update mri/operators/fourier/non_cartesian.py Comply to the pydocstyle guidelines Co-authored-by: Pierre-Antoine Comby <[email protected]> * Update mri/operators/fourier/non_cartesian.py comply to the pydocstyle guidelines Co-authored-by: Pierre-Antoine Comby <[email protected]> * Update mri/operators/fourier/non_cartesian.py comply to the pydocstyle guidelines Co-authored-by: Pierre-Antoine Comby <[email protected]> * Update mri/operators/fourier/non_cartesian.py comply to the pydocstyle guidelines Co-authored-by: Pierre-Antoine Comby <[email protected]> * Update mri/operators/fourier/cartesian.py comply to the pydocstyle guidelines Co-authored-by: Pierre-Antoine Comby <[email protected]> * Update mri/operators/fourier/cartesian.py comply to the pydocstyle guidelines Co-authored-by: Pierre-Antoine Comby <[email protected]> * Update mri/operators/fourier/cartesian.py Minor typo Co-authored-by: Pierre-Antoine Comby <[email protected]> * Update mri/operators/fourier/cartesian.py Minor fix Co-authored-by: Pierre-Antoine Comby <[email protected]> * Update mri/operators/fourier/cartesian.py Minor fix Co-authored-by: Pierre-Antoine Comby <[email protected]> * Update mri/operators/fourier/non_cartesian.py * Update mri/operators/fourier/non_cartesian.py * Update mri/operators/fourier/non_cartesian.py * Update mri/operators/fourier/non_cartesian.py * Update mri/reconstructors/base.py Co-authored-by: Pierre-Antoine Comby <[email protected]> * Update mri/reconstructors/base.py Co-authored-by: Pierre-Antoine Comby <[email protected]> * Update mri/reconstructors/base.py Co-authored-by: Pierre-Antoine Comby <[email protected]> * Apply suggestions from code review Co-authored-by: Pierre-Antoine Comby <[email protected]> * Update mri/reconstructors/base.py Co-authored-by: Pierre-Antoine Comby <[email protected]> * [doc] s/: np\./: numpy\./g * [doc] s/""" /"""/g * [doc][fourier] Constructor parameters in class docstring * [doc] s/: ndarray/: numpy.ndarray/g --------- Co-authored-by: Zaineb AMOR <[email protected]> Co-authored-by: Philippe Ciuciu <[email protected]> Co-authored-by: Guillaume DAVAL FREROT <[email protected]> Co-authored-by: Pierre-Antoine Comby <[email protected]> Co-authored-by: Daval-G <[email protected]> Co-authored-by: Pierre-antoine Comby <[email protected]> * Update versioning * Updated codes * Fix one example * Cartesian recon * Fix Density comp * Added density compensation and gpunufft stuff * refactored all examples * Fix GL * Update setup.py * Update setup.py * Update setup.py * Fix density compensation comparison module * Added density compensation module * Create gpu_requirements * Rename gpu_requirements to gpu_requirements.txt * Update README.txt * Update gpu_requirements.txt * Update gpu_requirements.txt * Update setup.py * Added auto tuto * Final large iteration * Move back to GPU for speed * Minor bug fix * Checking done * fix versioning * remove * Fix auto thresh * Fix grid search * Update modopt --------- Co-authored-by: Philippe Ciuciu <[email protected]> Co-authored-by: Pierre-antoine Comby <[email protected]> Co-authored-by: Pierre-Antoine Comby <[email protected]> Co-authored-by: Daval-G <[email protected]> Co-authored-by: Zaineb AMOR <[email protected]> Co-authored-by: Guillaume DAVAL FREROT <[email protected]> Co-authored-by: GILIYAR RADHAKRISHNA Chaithya <[email protected]>
CEA-COSMIC · Jan 19, 2024 · d32ebdf · d32ebdf
1 parent 7e65415
commit d32ebdf
Show file tree

Hide file tree

Showing 59 changed files with 1,771 additions and 1,553 deletions.
diff --git a/.github/workflows/ci-build.yml b/.github/workflows/ci-build.yml
@@ -47,7 +47,7 @@ jobs:
           python -m pip install git+https://github.com/CEA-COSMIC/ModOpt.git@develop
           python -m pip install git+https://github.com/CEA-COSMIC/pysap.git@develop
           python -m pip install git+https://github.com/AGrigis/pysphinxdoc.git
-          python -m pip install coverage nose pytest pytest-cov pycodestyle twine pytest-xdist
+          python -m pip install coverage nose pytest pytest-cov pycodestyle pydocstyle twine pytest-xdist
           python -m pip install pynfft2
           python -m pip install --upgrade .
 
@@ -57,7 +57,8 @@ jobs:
         run: |
           pycodestyle mri --ignore="E121,E123,E126,E226,E24,E704,E402,E731,E722,E741,W503,W504,W605"
           pycodestyle examples --ignore="E121,E123,E126,E226,E24,E704,E402,E731,E722,E741,W503,W504,W605"
-
+          pydocstyle  mri --convention=numpy
+          pydocstyle  examples --convention=numpy
       - name: Run Tests
         shell: bash -l {0}
         run: |

diff --git a/doc/refs.bib b/doc/refs.bib
@@ -34,3 +34,24 @@ @article{man1997
     year = {1997},
     pages = {785-792},
 }
+
+@article{Pruessmann1999,
+  title={SENSE: Sensitivity encoding for fast MRI},
+  author={Klaas Paul Pruessmann and Markus Weiger and Markus B. Scheidegger and Peter Boesiger},
+  journal={Magnetic Resonance in Medicine},
+  year={1999},
+  volume={42},
+}
+@inproceedings{Donoho1994,
+  address         = {Baltimore, MD, USA},
+  title           = {Threshold selection for wavelet shrinkage of noisy data},
+  ISBN            = {978-0-7803-2050-5},
+  url             = {http://ieeexplore.ieee.org/document/412133/},
+  DOI             = {10.1109/IEMBS.1994.412133},
+  booktitle       = {Proceedings of 16th Annual International Conference of the
+                  IEEE Engineering in Medicine and Biology Society},
+  publisher       = {IEEE},
+  author          = {Donoho, D.L. and Johnstone, I.M.},
+  year            = 1994,
+  pages           = {A24–A25}
+}
diff --git a/examples/CalibrationlessReconstruction_GL_Cartesian.py b/examples/CalibrationlessReconstruction_GL_Cartesian.py
@@ -1,6 +1,6 @@
 """
 Cartesian Calibrationless Reconstruction using GroupLASSO Regularizer
-=====================================================================
+==============================================================
 
 Author: Chaithya G R
 
@@ -13,32 +13,36 @@
 We use the toy datasets available in pysap, more specifically a 2D parallel MRI
 brain slice on 32 channels and the acquisition cartesian scheme.
 """
-
+# %%
 # Package import
 from mri.operators import FFT, WaveletN
-from mri.operators.utils import convert_mask_to_locations
 from mri.reconstructors import CalibrationlessReconstructor
-import pysap
 from pysap.data import get_sample_data
 
 # Third party import
-from modopt.math.metrics import ssim
 from modopt.opt.proximity import GroupLASSO
+from modopt.math.metrics import ssim
 import numpy as np
+import matplotlib.pyplot as plt
 
-
+# %%
 # Loading input data
-cartesian_ref_image = get_sample_data('2d-pmri')
-image = pysap.Image(data=np.sqrt(np.sum(cartesian_ref_image.data**2, axis=0)))
+cartesian_ref_image = get_sample_data('2d-pmri').data
+image = np.linalg.norm(cartesian_ref_image, axis=0)
 # Obtain MRI cartesian mask
-mask = get_sample_data("cartesian-mri-mask")
-kspace_loc = convert_mask_to_locations(mask.data)
+mask = get_sample_data("cartesian-mri-mask").data
 
+# %%
 # View Input
-# image.show()
-# mask.show()
+plt.subplot(1, 2, 1)
+plt.imshow(np.abs(image), cmap='gray')
+plt.title("MRI Data")
+plt.subplot(1, 2, 2)
+plt.imshow(mask, cmap='gray')
+plt.title("K-space Sampling Mask")
+plt.show()
 
-#############################################################################
+# %%
 # Generate the kspace
 # -------------------
 #
@@ -47,18 +51,19 @@
 # We then reconstruct the zero order solution as a baseline
 
 # Get the locations of the kspace samples and the associated observations
-fourier_op = FFT(samples=kspace_loc, shape=image.shape,
+fourier_op = FFT(mask=mask, shape=image.shape,
                  n_coils=cartesian_ref_image.shape[0])
 kspace_obs = fourier_op.op(cartesian_ref_image)
 
-# Zero Filled reconstruction
-zero_filled = fourier_op.adj_op(kspace_obs)
-image_rec0 = pysap.Image(data=np.sqrt(np.sum(np.abs(zero_filled)**2, axis=0)))
-# image_rec0.show()
-base_ssim = ssim(image_rec0, image)
-print('The Base SSIM is : ' + str(base_ssim))
+# %%
+# Zero order solution
+zero_soln = np.linalg.norm(fourier_op.adj_op(kspace_obs), axis=0)
+base_ssim = ssim(zero_soln, image)
+plt.imshow(np.abs(zero_soln), cmap='gray')
+plt.title('Zero Order Solution : SSIM = ' + str(np.around(base_ssim, 2)))
+plt.show()
 
-#############################################################################
+# %%
 # FISTA optimization
 # ------------------
 #
@@ -71,6 +76,7 @@
     nb_scale=4,
     n_coils=cartesian_ref_image.shape[0],
 )
+coeffs = linear_op.op(cartesian_ref_image)
 regularizer_op = GroupLASSO(weights=6e-8)
 # Setup Reconstructor
 reconstructor = CalibrationlessReconstructor(
@@ -80,12 +86,15 @@
     gradient_formulation='synthesis',
     verbose=1,
 )
-x_final, costs, metrics = reconstructor.reconstruct(
+# %%
+# Run the FISTA reconstruction and view results
+image_rec, costs, metrics = reconstructor.reconstruct(
     kspace_data=kspace_obs,
     optimization_alg='fista',
-    num_iterations=300,
+    num_iterations=100,
 )
-image_rec = pysap.Image(data=np.sqrt(np.sum(np.abs(x_final)**2, axis=0)))
+image_rec = np.linalg.norm(image_rec, axis=0)
 recon_ssim = ssim(image_rec, image)
-print('The Reconstruction SSIM is : ' + str(recon_ssim))
-# image_rec.show()
+plt.imshow(np.abs(image_rec), cmap='gray')
+plt.title('Iterative Reconstruction : SSIM = ' + str(np.around(recon_ssim, 2)))
+plt.show()
diff --git a/examples/CalibrationlessReconstruction_OWL_Cartesian.py b/examples/CalibrationlessReconstruction_OWL_Cartesian.py
@@ -13,31 +13,35 @@
 We use the toy datasets available in pysap, more specifically a 2D parallel MRI
 brain slice on 32 channels and the acquisition cartesian scheme.
 """
-
+# %%
 # Package import
 from mri.operators import FFT, WaveletN, OWL
-from mri.operators.utils import convert_mask_to_locations
 from mri.reconstructors import CalibrationlessReconstructor
-import pysap
 from pysap.data import get_sample_data
 
 # Third party import
 from modopt.math.metrics import ssim
 import numpy as np
+import matplotlib.pyplot as plt
 
-
+# %%
 # Loading input data
-cartesian_ref_image = get_sample_data('2d-pmri')
-image = pysap.Image(data=np.sqrt(np.sum(cartesian_ref_image.data**2, axis=0)))
+cartesian_ref_image = get_sample_data('2d-pmri').data
+image = np.linalg.norm(cartesian_ref_image, axis=0)
 # Obtain MRI cartesian mask
-mask = get_sample_data("cartesian-mri-mask")
-kspace_loc = convert_mask_to_locations(mask.data)
+mask = get_sample_data("cartesian-mri-mask").data
 
+# %%
 # View Input
-# image.show()
-# mask.show()
+plt.subplot(1, 2, 1)
+plt.imshow(np.abs(image), cmap='gray')
+plt.title("MRI Data")
+plt.subplot(1, 2, 2)
+plt.imshow(mask, cmap='gray')
+plt.title("K-space Sampling Mask")
+plt.show()
 
-#############################################################################
+# %%
 # Generate the kspace
 # -------------------
 #
@@ -46,18 +50,19 @@
 # We then reconstruct the zero order solution as a baseline
 
 # Get the locations of the kspace samples and the associated observations
-fourier_op = FFT(samples=kspace_loc, shape=image.shape,
+fourier_op = FFT(mask=mask, shape=image.shape,
                  n_coils=cartesian_ref_image.shape[0])
 kspace_obs = fourier_op.op(cartesian_ref_image)
 
-# Zero Filled reconstruction
-zero_filled = fourier_op.adj_op(kspace_obs)
-image_rec0 = pysap.Image(data=np.sqrt(np.sum(np.abs(zero_filled)**2, axis=0)))
-# image_rec0.show()
-base_ssim = ssim(image_rec0, image)
-print('The Base SSIM is : ' + str(base_ssim))
+# %%
+# Zero order solution
+zero_soln = np.linalg.norm(fourier_op.adj_op(kspace_obs), axis=0)
+base_ssim = ssim(zero_soln, image)
+plt.imshow(np.abs(zero_soln), cmap='gray')
+plt.title('Zero Order Solution : SSIM = ' + str(np.around(base_ssim, 2)))
+plt.show()
 
-#############################################################################
+# %%
 # FISTA optimization
 # ------------------
 #
@@ -86,11 +91,15 @@
     gradient_formulation='synthesis',
     verbose=1,
 )
-x_final, costs, metrics = reconstructor.reconstruct(
+# %%
+# Run the FISTA reconstruction and view results
+image_rec, costs, metrics = reconstructor.reconstruct(
     kspace_data=kspace_obs,
     optimization_alg='fista',
-    num_iterations=200,
+    num_iterations=100,
 )
-image_rec = pysap.Image(data=np.sqrt(np.sum(np.abs(x_final)**2, axis=0)))
+image_rec = np.linalg.norm(image_rec, axis=0)
 recon_ssim = ssim(image_rec, image)
-print('The Reconstruction SSIM is : ' + str(recon_ssim))
+plt.imshow(np.abs(image_rec), cmap='gray')
+plt.title('Iterative Reconstruction : SSIM = ' + str(np.around(recon_ssim, 2)))
+plt.show()
diff --git a/examples/GPU_Examples/NonCartesian_gpuNUFFT_DensityCompensation.py b/examples/GPU_Examples/NonCartesian_gpuNUFFT_DensityCompensation.py