This page contains the benchmarks for the CASToR 3.xx versions. Each benchmark consists in a dataset and a reference image that is a reconstruction of this dataset. A script (provided for both Unix and Windows systems) can be used to reconstruct the dataset with your own CASToR installation and check the consistency between the reconstructed image and the reference image. The consistency check is performed by a program that analyzes the differences of voxels' values and sends a warning when beyond a preset value.
These benchmarks can be a good starting point to have a quick look at the input file formats as well as the available options.
This page also contains some examples which will just perform a reconstruction of simulated/projection data. The aim of these files is to show how to configure CASToR for different uses.
PET benchmarks:
Histogram
This benchmark is based on a real acquisition of a whole-body 18F-FDG TOF-PET data set acquired on a SIGNA PET/MR system. The histogram data file includes azimuthal and axial compressions and contains all corrections (normalization, random, scatter and attenuation).
List-mode
This benchmark is based on a real acquisition of a whole-body 18F-FDG TOF-PET data set acquired on a SIGNA PET/MR system. The list-mode data file contains all corrections (normalization, random, scatter and attenuation). A normalization data file is provided to precompute the sensitivity image. The benchmark will test the consistency of both the reconstructed and the sensitivity images.
SPECT benchmark:
Histogram
This benchmark is based on a real acquisition of a Jaszczak phantom data set filled with 99mTc, using a dual-head SPECT/CT system with a parallel collimator. An attenuation map is provided in order to perform attenuation correction.
CT benchmark:
Histogram
This benchmark is based on a simulated 52 keV acquisition of a digital phantom data set, using a cone-beam CT (CBCT) system. The simulated phantom is composed of a cylinder of water containing inserts made of lung, blood, brain and polystyrene materials, and a central vacuum insert.
Dynamic examples:
PET Dynamic (Patlak, Spectral, Frame by frame)
This example uses data generated from the Zubal head phantom to generate standard frame by frame, patlak and spectral reconstruction of a simulated dynamic PET histogram dataset. The PET system used in this example contains only 1 ring in order to decrease data size.
PET dmodel-rmotion (5D Patlak with rigid motion deformation)
This example uses data generated from the Zubal head phantom to perform:
- 5D reconstruction incorporation the Patlak model and rigid motion correction for random motion (timestamp-based motion correction).
The dataset is a List-mode generated from analytical projection of Zubal digital phantom with attenuation and noise
The activity in voxels represent typical FDG time activity curves.
The dynamic protocol starts at 2100 seconds, to reconstruct 5 frames of 300s each.
5 motion steps are simulated at 2280s, 2460s, 3060s, 3180s and 3480s.
- Post-reconstruction voxelwise kinetic fitting with Patlak from 3D (frame by frame) pre-reconstructed images.
The dataset from which the images were reconstructed is the same than for the 5D reconstruction above.