The SIMPET project is intended to allow to setup and launch MC simulation on a simple way. It provides functionalities to:
- Extract simple activity and attenuation maps from PET/MR images.
- Apply the BrainViset procedure to obtain realistic Activity and Attenuation maps.
- Run Analytic simulations using STIR simulation procedure and MC simulation using SimSET.
- Install Git LFS.
- Clone the repository by adding the
--recurse-submodulesflag:
git clone --recurse-submodules --branch develop https://github.com/txusser/simpet.git
- Install python 3.9. If you use apt (most Debian based distros use it):
sudo add-apt-repository ppa:deadsnakes/ppa
sudo apt install python3.9
- Create a python virtual environment from python 3.9, there are several options for this (we recommend Conda):
- venv.
- Conda.
- virtualenv.
You can isntall miniconda in Linux with the following code:
mkdir -p ~/miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm -rf ~/miniconda3/miniconda.sh
Then restart your shell:
~/miniconda3/bin/conda init bash
~/miniconda3/bin/conda init zsh
- Activate the virtual environment and install the requirements with:
pip install -r requirements.txt
Sometimes, even activating the virtual environemnt, the shell will use the wide system pip, you can check it with which pip (in Linux). In that case, you may want to locate the python interpreter of your virtual environment or an alias (most of the times python3.9) then run:
/path/to/your/virtual-environtment/bin/python3.9 -m pip install -r requirements.txt
- Install the project with (we recommend to be an user with root privileges):
make install
- Decompress dummy data with:
make dummy-data
- Activate your virtual environment and run a test with the simulation launcher:
python3.9 scripts/experiment.py
To make the most of this version of the project is strongly recommended to be familiar with facebook-hydra. The configuration of a given simulation (and reconstruction) is split into 3 groups: global configuration, params, configuration and scanner configuration (examples given below). The scanner group is a subgroup of params group and params is a subgroup of the global configuration.
- Example of
globalconfiguration. - Example of
paramsconfiguration group. - Example of
scannerconfiguration group.
This approach allows the user to have several configuration groups and switch between them at no cost. For example, given the following tree of configurations (each subgroup has a "production" configuration and a "test" configuration):
.
└── configs/
├── config_prod.yaml # global configuration group
├── config_test.yaml # global configuration group
└── params/
├── params_prod.yaml
├── params_test.yaml
└── scanner/
├── siemens.yaml
└── discovery.yaml
Using facebook-hydra override syntax and the experiment.py launcher, switching between configurations is trivial:
# Running test configuration with Siemens scanner
python scripts/experiment.py --config-name config_test params/scanner=siemens
# Running prod configuration with test `params` and Discovery scanner
python scripts/experiment.py --config-name config_prod params=params_test params/scanner=discovery
# Running prod configuration with test `params` and Discovery scanner but doing only reconstruction
python scripts/experiment.py --config-name config_prod params=params_test params/scanner=discovery params.do_simulation=0
# Running prod configuration with test `params` and Discovery scanner, doing only reconstruction and overriding the scanner radius
python scripts/experiment.py --config-name config_prod params=params_test params/scanner=discovery params.do_simulation=0 params.scanner.scanner_radius=35
You can perform whole body simulations following the same logic described in the last section. You may want to add z_min and z_max parameters to the params configuration group. Here you can find the configuration templates:
Then you can launch an experiment with the experiment_wholebody.py launcher:
python3.9 scripts/experiment_wholebody.py --config-name <your_config_name>
Even add z_min and z_max on the fly:
python3.9 scripts/experiment_wholebody.py --config-name <your_config_name> +params.z_min=29 +params.z_max=89
Or override them:
python3.9 scripts/experiment_wholebody.py --config-name <your_config_name> params.z_min=29 params.z_max=89
BrainVISET is an iterative algorithm that allows the generation of activity and attenuation maps from high-resolution CT and MRI images. To run BrainVISET you will need SPM12 and MATLAB MCR. Ensure that the configuration keys matlab_mcr_path (path to MATLAB MCR) and spm_path (path to SPM12) are well set in your configs file. In addition, you must also specify the names (with extension) of the CT, MRI and PET images in your configs/params file (keys ct_image, mri_image and pet_image). Then, you can run BrainVISET using the experiment_brainviset.py launcher:
python3.9 scripts/experiment_brainviset.py --config-name <your_config_name>
Even change image names on the fly (or other parameters):
python3.9 scripts/experiment_brainviset.py --config-name <your_config_name> params.ct_image="my_ct.nii" params.mri_image="my_mri.nii" params.pet_image="my_pet.nii"
GENERAL NOTES ON USAGE:
- Monte Carlo simulations usually take a lot of space. If you want to change the default Data and Results directories you can do so in the in the
globalconfiguration group or using facebook-hydra CLI overriding syntax. - The current version of the project can produce only sinograms and LM data using SimSET. Rest of the stuff is coming soon.
- The repository has only been tested on Ubuntu 22.04.1 LTS.
- dir_stir: STIR directory (you should not worry about this).
- dir_simset: SimSET directory (you should not worry about this).
- matlab_mcr_path: path to MATLAB MCR, needed for BrainVISET.
- spm_path: path to SPM12, needed for BrainVISET.
- dir_data_path: path to directory with patients or subjects.
- dir_results_path: path to store the results of the simulation + reconstruction.
- interactive_mode: if set to one you may see some prompts to prevent you to overwrite existing data
- stratification: SimSET variance technique reduction. We recommend setting it to "true"
- forced_detection: SimSET variance technique reduction. We recommend setting it to "true".
- forced_non_absortion: We recommend setting it to "true". We recommend setting it to "true".
- acceptance_angle: SimSET acceptance angle. We recommend setting it to 90.0 but it may need fine tunning for your case.
- positron_range: whether to simulate positron range or not with SimSET. We recommend setting it to "true".
- non_colinearity: whether to simulate non-colinearity of photons with SimSET. We recommend setting it to "true".
- minimum_energy: minimum energy threshold. We recommend setting it to 350.0.
- weight_window_ratio: we recommend setting it to 1.0.
- point_source_voxels: we recommend setting it to "false".
- coherent_scatter_object: we recommend setting it to "false".
- coherent_scatter_detector: we recommend setting it to "false".
See SimSET documentation for further information.
- sim_type: one of SimSET, STIR or GATE (only SimSET working).
- do_simulation: 1 or 0, whether to perform simulation (1) or not (0). This is useful for deactivating simulation when only reconstruction is needed.
- do_reconstruction: 1 or 0, whether to perform reconstruction (1) or not (0).
- divisions: number os subprocesses for parallel simulation.
- scanner_name: name of the PET system.
- model_type: we recommend setting it to "cylindrical".ç
- patient_dirname: name of the subdirectory in dir_data_path where phantoms are located.
- output_dir: name of the attenuation map file (with extension).
- act_map: name of the subdirectory in dir_results_path where the results will be stored.
- att_map: name of the activity map file (with extension).
- center_slice: the slice number to be placed on the center of the scanner. If 0, automatically, the half of the slices will be calculated and used.
- z_min: start of the acquisition, slice.
- z_max: end of the acquisition, slice. We calculate the number of beds based on (z_max - z_min) and the scanner FOV.
- mri_image: MRI filename with extension.
- ct_image: CT filename with extension.
- pet_image: PET filename with extension.
- maximumIteration: maximum number of iterations when using BrainVISET.
- total_dose: dose in mCi.
- simulation_time: time to be simulated in s.
- sampling_photons: set to 0 to avoid importance sampling. We recommend the use of importance sampling and setting this parameter to 20000000 as a starting point.
- photons: set to 0 to do a realistic noise simulation.
- add_randoms: 1 activate randoms simulation (will force sampling_photons=0 and photons=0). It can kill the process is there is not enough RAM.
- phglistmode: history Files from the phg module (needed for LM reconstruction, potentially very big).
- detlistmode: history Files from the detector module (you need this for adding randoms, if add_randoms=1 will be forced).
- scanner_name: name of the scanner.
- simset_material: material of the scanner, see SimSET for available materials.
- average_doi: depth of interaction.
- scanner_radius: radius of the scanner in cm.
- num_rings: number of rings of the scanner.
- axial_fov: axial FOV of the scanner in cm.
- z_crystal_size: size of the crystal in the axial direction in cm.
- transaxial_crystal_size: size of the cristal in the transaxial size in cm.
- crystal_thickness: thickness of the crystal in cm.
- energy_resolution: FWHM in keV it will be divided by reference energy (511 keV).
- min_energy_window: min energy threshold in keV.
- max_energy_window: max energy threshold in keV.
- num_aa_bins: number of views (half of the number of detectors per ring).
- num_td_bins: number of radial bins.
- coincidence_window: used only for randoms simulation, in ns.
- psf_value: sinogram preprocessing parameter.
- add_noise: sinogram preprocessing parameter.
- analytical_att_correction: performed by SimSET calcattenuation.
- stir_recons_att_corr: performed in STIR by entering the att image as a normalization map.
- analytic_scatt_corr_factor: 0.15 will remove 85% of scatter, 0 will remove scatter.
- stir_scatt_corr_smoothing: will use smoothed SimSET scatter as additive_sinogram.
- analytic_randoms_corr_factor: 0.15 will remove 85% of scatter, 0 will remove randoms.
- stir_randoms_corr_smoothing: will include smoothed SimSET randoms in the additive_sinogram.
- recons_type: one of OSEM3D, OSEM2D, FBP2D, FBP3D.
- max_segment: maximum Ring difference on the reconstruction.
- zoomFactor: relation between sinogram bin size and X, Y voxel size.
- xyOutputSize: output matrix X,Y size in reconstruction.
- zOutputSize: output marix Z size in reconstruction.
- numberOfSubsets: number of subsets in reconstruction.
- numberOfIterations: number of iterations in iterative reconstruction.
- savingInterval: interval that to save intermediate subiterations images.
- inter_iteration_filter: 1 for activation, 0 for deactivation.
- subiteration_interval:
- x_dir_filter_FWHM: postprocessing filter X size in mm.
- y_dir_filter_FWHM: postprocessing filter Y size in mm.
- z_dir_filter_FWHM: postprocessing filter Z size in mm.
- Jesús Silva-Rodríguez
- Pablo Aguiar
- Aida Ninyerola-Baizan
- Jeremiah Poveda
- Francisco Javier López-González
- Nikos Efthimiou
- Arnau Farre
You may find more information about STIR and SimSET parameters in their official documentation.
[1] SimSET.
[2] STIR.