Skip to content

Use more accurate adjoint weighting for 2D geometries in ASTRA #2135

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wjp wants to merge 3 commits into
base: master
Choose a base branch
from
Open

Use more accurate adjoint weighting for 2D geometries in ASTRA #2135

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
9968525
Switch to ASTRA direct_FP3D/BP3D interface
wjp Apr 8, 2025
51cc36d
Add self to contributors
wjp Apr 8, 2025
ec987cf
Use 2d_weighting for ASTRA adjoint when in 2D mode
wjp Apr 8, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions NOTICE.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,6 +35,7 @@ Institutions Key:
12 - Australian e-Health Research, Australia
13 - KU Leuven
14 - Independent Contributor
15 - Universiteit Leiden

CIL Developers in date order:
Edoardo Pasca (2017 – present) - 1
Expand Down Expand Up @@ -72,6 +73,7 @@ Nicholas Whyatt (2024) - 1
Rasmia Kulan (2024) - 1
Emmanuel Ferdman (2025) - 14
Mariam Demir (2025) - 1
Willem Jan Palenstijn (2025) - 15

CIL Advisory Board:
Llion Evans - 9
Expand Down
93 changes: 19 additions & 74 deletions Wrappers/Python/cil/plugins/astra/processors/AstraBackProjector3D.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,10 +16,10 @@
# Authors:
# CIL Developers, listed at: https://github.com/TomographicImaging/CIL/blob/master/NOTICE.txt
from cil.framework import DataProcessor, ImageData
from cil.framework.labels import AcquisitionDimension, ImageDimension
from cil.framework.labels import AcquisitionDimension, AcquisitionType, ImageDimension
from cil.plugins.astra.utilities import convert_geometry_to_astra_vec_3D
import astra
from astra import astra_dict, algorithm, data3d
import astra.experimental
import numpy as np


Expand All @@ -44,16 +44,25 @@ def __init__(self,
kwargs = {
'volume_geometry' : volume_geometry,
'sinogram_geometry' : sinogram_geometry,
'proj_geom' : None,
'vol_geom' : None}
'projector_id' : None}

#DataProcessor.__init__(self, **kwargs)
super(AstraBackProjector3D, self).__init__(**kwargs)

self.set_ImageGeometry(volume_geometry)
self.set_AcquisitionGeometry(sinogram_geometry)

self.vol_geom, self.proj_geom = convert_geometry_to_astra_vec_3D(self.volume_geometry, self.sinogram_geometry)
vol_geom, proj_geom = convert_geometry_to_astra_vec_3D(self.volume_geometry, self.sinogram_geometry)

proj_cfg = astra.astra_dict('cuda3d')
proj_cfg['ProjectionGeometry'] = proj_geom
proj_cfg['VolumeGeometry'] = vol_geom
if AcquisitionType.DIM2 & self.sinogram_geometry.dimension:
proj_cfg['ProjectionKernel'] = '2d_weighting'
self.projector_id = astra.projector3d.create(proj_cfg)

def __del__(self):
astra.projector3d.delete(self.projector_id)

def check_input(self, dataset):

Expand Down Expand Up @@ -95,20 +104,15 @@ def process(self, out=None):
new_shape_ag = [self.sinogram_geometry.pixel_num_v,self.sinogram_geometry.num_projections,self.sinogram_geometry.pixel_num_h]
data_temp = DATA.as_array().reshape(new_shape_ag)

new_shape_ig = [self.volume_geometry.voxel_num_z,self.volume_geometry.voxel_num_y,self.volume_geometry.voxel_num_x]
new_shape_ig = [x if x>0 else 1 for x in new_shape_ig]

if out is None:
rec_id, arr_out = astra.create_backprojection3d_gpu(data_temp,
self.proj_geom,
self.vol_geom)
arr_out = np.zeros(new_shape_ig, dtype=np.float32)
else:
new_shape_ig = [self.volume_geometry.voxel_num_z,self.volume_geometry.voxel_num_y,self.volume_geometry.voxel_num_x]
new_shape_ig = [x if x>0 else 1 for x in new_shape_ig]
arr_out = out.as_array().reshape(new_shape_ig)

rec_id = astra.data3d.link('-vol', self.vol_geom, arr_out)
self.create_backprojection3d_gpu( data_temp, self.proj_geom, self.vol_geom, False, rec_id)

# delete the GPU copy
astra.data3d.delete(rec_id)
astra.experimental.direct_BP3D(self.projector_id, arr_out, data_temp)

arr_out = np.squeeze(arr_out)

Expand All @@ -117,62 +121,3 @@ def process(self, out=None):
else:
out.fill(arr_out)
return out

def create_backprojection3d_gpu(self, data, proj_geom, vol_geom, returnData=True, vol_id=None):

"""
Call to ASTRA to create a backward projection of an image (3D)

Parameters
----------

data : numpy.ndarray or int
Image data or ID.

proj_geom : dict
Projection geometry.

vol_geom : dict
Volume geometry.

returnData : bool
If False, only return the ID of the forward projection.

vol_id : int, default None
ID of the np array linked with astra.

Returns
-------

proj_geom : int or (int, numpy.ndarray)
If ``returnData=False``, returns the ID of the back projection. Otherwise returns a tuple containing the ID of the back projection and the back projection itself.
"""

if isinstance(data, np.ndarray):
sino_id = data3d.create('-sino', proj_geom, data)
else:
sino_id = data

if vol_id is None:
vol_id = data3d.create('-vol', vol_geom, 0)

cfg = astra_dict('BP3D_CUDA')
cfg['ProjectionDataId'] = sino_id
cfg['ReconstructionDataId'] = vol_id
alg_id = algorithm.create(cfg)
algorithm.run(alg_id)
algorithm.delete(alg_id)

if isinstance(data, np.ndarray):
data3d.delete(sino_id)

if vol_id is not None:
if returnData:
return vol_id, data3d.get_shared(vol_id)
else:
return vol_id
else:
if returnData:
return vol_id, data3d.get(vol_id)
else:
return vol_id
86 changes: 12 additions & 74 deletions Wrappers/Python/cil/plugins/astra/processors/AstraForwardProjector3D.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@
from cil.framework.labels import ImageDimension, AcquisitionDimension
from cil.plugins.astra.utilities import convert_geometry_to_astra_vec_3D
import astra
from astra import astra_dict, algorithm, data3d
import astra.experimental
import numpy as np

class AstraForwardProjector3D(DataProcessor):
Expand All @@ -43,16 +43,20 @@ def __init__(self,
kwargs = {
'volume_geometry' : volume_geometry,
'sinogram_geometry' : sinogram_geometry,
'proj_geom' : None,
'vol_geom' : None,
'projector_id' : None
}

super(AstraForwardProjector3D, self).__init__(**kwargs)

self.set_ImageGeometry(volume_geometry)
self.set_AcquisitionGeometry(sinogram_geometry)

self.vol_geom, self.proj_geom = convert_geometry_to_astra_vec_3D(self.volume_geometry, self.sinogram_geometry)
vol_geom, proj_geom = convert_geometry_to_astra_vec_3D(self.volume_geometry, self.sinogram_geometry)
self.projector_id = astra.create_projector('cuda3d', proj_geom, vol_geom)

def __del__(self):
astra.projector3d.delete(self.projector_id)


def check_input(self, dataset):

Expand Down Expand Up @@ -97,19 +101,14 @@ def process(self, out=None):

data_temp = IM.as_array().reshape(new_shape_ig)

new_shape_ag = [self.sinogram_geometry.pixel_num_v,self.sinogram_geometry.num_projections,self.sinogram_geometry.pixel_num_h]

if out is None:
sinogram_id, arr_out = astra.create_sino3d_gpu(data_temp,
self.proj_geom,
self.vol_geom)
arr_out = np.zeros(new_shape_ag, dtype=np.float32)
else:
new_shape_ag = [self.sinogram_geometry.pixel_num_v,self.sinogram_geometry.num_projections,self.sinogram_geometry.pixel_num_h]
arr_out = out.as_array().reshape(new_shape_ag)

sinogram_id = astra.data3d.link('-sino', self.proj_geom, arr_out)
self.create_sino3d_gpu(data_temp, self.proj_geom, self.vol_geom, False, sino_id=sinogram_id)

#clear the memory on GPU
astra.data3d.delete(sinogram_id)
astra.experimental.direct_FP3D(self.projector_id, data_temp, arr_out)

arr_out = np.squeeze(arr_out)

Expand All @@ -118,64 +117,3 @@ def process(self, out=None):
else:
out.fill(arr_out)
return out

def create_sino3d_gpu(self, data, proj_geom, vol_geom, returnData=True, gpuIndex=None, sino_id=None):
"""
Call to ASTRA to create a forward projection of an image (3D)

Parameters
----------

data : numpy.ndarray or int
Image data or ID.

proj_geom : dict
Projection geometry.

vol_geom : dict
Volume geometry.

returnData : bool
If False, only return the ID of the forward projection.

gpuIndex : int, optional
Optional GPU index.

Returns
-------

proj_geom : int or (int, numpy.ndarray)
If ``returnData=False``, returns the ID of the forward projection. Otherwise returns a tuple containing the ID of the forward projection and the forward projection itself.

"""


if isinstance(data, np.ndarray):
volume_id = data3d.create('-vol', vol_geom, data)
else:
volume_id = data
if sino_id is None:
sino_id = data3d.create('-sino', proj_geom, 0)

algString = 'FP3D_CUDA'
cfg = astra_dict(algString)
if not gpuIndex==None:
cfg['option']={'GPUindex':gpuIndex}
cfg['ProjectionDataId'] = sino_id
cfg['VolumeDataId'] = volume_id
alg_id = algorithm.create(cfg)
algorithm.run(alg_id)
algorithm.delete(alg_id)

if isinstance(data, np.ndarray):
data3d.delete(volume_id)
if sino_id is None:
if returnData:
return sino_id, data3d.get(sino_id)
else:
return sino_id
else:
if returnData:
return sino_id, data3d.get_shared(sino_id)
else:
return sino_id