A phantom generation package for cone beam CT geometries.
This generator analytically computes a circular cone beam projection of a configurable number of (potentially overlapping) solid spheres of constant density and varying random radius.
It randomly places the centers of the spheres in a cube of configurable size. Where the spheres overlap, their linear attenuation coefficients are summed.
The randomly generated ball positions and radii are stored for later re-use, as well as the used cone beam geometry (as an ASTRA projection geometry).
Cone Balls is GPU accelerated and requires CUDA.
- Free software: GNU General Public License v3
- Documentation is currently incomplete.
It takes a few steps to setup Cone Balls on your machine. We recommend installing Anaconda package manager for Python 3.
Simply install with:
conda install -c aahendriksen -c pytorch -c astra-toolbox/label/dev -c conda-forge cone_balls
To install Cone Balls, simply clone this GitHub project. Go to the cloned directory and run PIP installer:
git clone https://github.com/ahendriksen/cone_balls.git
cd cone_balls
conda install -c astra-toolbox/label/dev -c pytorch -c conda-forge cudatoolkit=10.0 astra-toolbox pytorch torchvision tifffile
pip install -e .
Cone balls is a command-line utility. It creates a directory with tiff files.
$ cone_balls --help
Usage: cone_balls [OPTIONS] COMMAND [ARGS]...
Options:
--help Show this message and exit.
Commands:
bench Time the generation of cone_balls
generate cone_balls generates ball phantoms for cone beam geometries
$ cone_balls generate --help
Usage: cone_balls generate [OPTIONS] DIR
generate generates cone-beam projections of ball phantoms
By default - 100 balls are randomly generated - 1500 projections are
computed on a 700 x 700 detector with pixel size 1.0 x 1.0 - The source-
object distance and the source-detector distance are 700.0, meaning that
the detector is centered on the origin and rotates through the object.
Options:
--num_balls INTEGER Number of balls to generate.
--ball_limit INTEGER The maximal distance from the origin of a
ball
--num_angles INTEGER Number of angles.
--det_row_count INTEGER Detector row count.
--det_col_count INTEGER Detector column count.
--pixel_size FLOAT The detector pixel size.
--SOD FLOAT The source object distance.
--SDD FLOAT The source detector distance.
--interactive / --no-interactive
Show geometry and resulting projection
images
--ball_spec FILE
--help Show this message and exit.
$ cone_balls foam --help
Usage: cone_balls foam [OPTIONS] DIR
foam generates cone-beam projections of a foam ball phantom
The foam ball has a radius of 0.5 and is centered on the origin. Bubbles
can be removed from this foam phantom. The location and size of these
bubbles can either be supplied using the --ball_spec option, or randomly
generated.
Options:
--num_balls INTEGER Number of balls to generate.
--ball_limit FLOAT The maximal distance from the origin of a
ball
--num_angles INTEGER Number of angles.
--det_row_count INTEGER Detector row count.
--det_col_count INTEGER Detector column count.
--pixel_size FLOAT The detector pixel size.
--SOD FLOAT The source object distance.
--SDD FLOAT The source detector distance.
--Z FLOAT The Z-offset of source and detector.
--interactive / --no-interactive
Show geometry and resulting projection
images
--ball_spec FILE
--help Show this message and exit.
$ cone_balls bench --help
Usage: cone_balls bench [OPTIONS]
Time the generation of cone_balls
Options:
--num_balls INTEGER Number of balls to generate.
--num_angles INTEGER Number of angles.
--det_pix_count INTEGER Detector column count.
--interactive / --no-interactive
Show geometry and resulting projection
images
--help Show this message and exit.
First generate projection data using cone_balls:
mkdir -p example
cone_balls generate --num_balls 10 --num_angles 50 example
Start a python console and run:
import astra
import astra.experimental
import numpy as np
from pathlib import Path
import tifffile
import pickle
# Load data
tiffs = [tifffile.imread(str(p)) for p in sorted(Path("example/").glob("*.tif"))]
# Put projection data in [Y, Angle, X] order
proj_data = np.array(tiffs).swapaxes(0, 1)
# Load projection geometry
pg = pickle.load(open("example/astra_geometry.pickle", "rb"))
# Create astra projection data
p_id = astra.data3d.create('-proj3d', pg, proj_data)
vg = {'option': {'WindowMinX': -350, 'WindowMaxX': 350,
'WindowMinY': -350, 'WindowMaxY': 350,
'WindowMinZ': -350, 'WindowMaxZ': 350},
'GridRowCount': 250,
'GridColCount': 250,
'GridSliceCount': 250
}
v_id = astra.data3d.create('-vol', vg)
projector = astra.create_projector("cuda3d", pg, vg)
astra.experimental.accumulate_FDK(projector, v_id, p_id)
# Display volume data
import pyqtgraph as pq
app = pq.mkQApp()
pq.image(astra.data3d.get_shared(v_id))
app.exec_()The display of data might require that you install pyqtgraph. This
package can be installed by executing
conda install pyqtgraph- Allard Hendriksen - Initial work
See also the list of contributors who participated in this project.
Contributions are always welcome. Please submit pull requests against the master branch.
If you have any issues, questions, or remarks, then please open an issue on GitHub.
This project is licensed under the GNU General Public License v3 - see the LICENSE.md file for details.
- To Willem Jan Palenstijn for useful advice and discussion.