shifeng1981 / dpm-g4cpp

dpm-g4cpp is a C++ prototype of the Dose Planning Method (DPM): fast, accurate dose Monte Carlo simulation for photon, electron radiotherapy treatment planning. As a clean C++ implementation of the simulation, that doesn't contain any non-deterministic loops in the interaction sampling, it serves as a more suitable basis for GPU/FPGA implementations.

The dpm-g4cpp R&D

Motivation

The motivations behind this project are:

• provide a clean, well documented C++ implementation of the well known, highly specialised Dose Planning Method (DMP)[1] Monte Carlo simulation that can be a basis for further GPU/FPGA developments
• in order to make the original DPM like simulation more suitable for these hardware, eliminate all possible non-deterministic loops from the simulation, especially those related to the final state generation of the individual interactions
• this project also serves as testbed for studying the possible benefits of the related treatment of the description of e-/e+ Multiple Coulomb Scattering (MSC) over the traditional models
• especially that the DPM like handling of MSC is very efficient and stays accurate (without paying any computing performance penalties) even in the case of highly granular geometries
• this latter is due to the fact that this simulation algorithm was designed and optimised for efficient and accurate modelling of particle transport through multiple, small geometry voxels

Description

dpm-g4cpp is a C++ prototype of the known Dose Planning Method (DMP)[1], fast and accurate Monte Carlo simulation. DMP is highly optimised for photon and electron radiotherapy treatment planning dose calculations based on a data extraction and simulation phase.

These two phases are also preserved and clearly separated in the dpm-g4cpp version:

• dpm_GenerateData : can be used to generate and write into files all the data that are needed for the later simulation phase. Since some of the data (the integrated data such as IMFP, dE/dx, etc.) are generated based on Geant4, this part requires a recent Geant4 (version >= 10.7.p01) installed on the system.
• dpm_Simulate : can be used to perform the dose simulation relying on the previously generated data files. One of the main difference between the original DPM and dpm-g4cpp that this latter simulation dose not contain any non-deterministic loops (rejection loops) during the interaction computation including even the sampling of angular deflections along a given path length according to the Goudsmit Saunderson angular distributions or the emitted bremsstralung photon energy according to the numerical Seltzer-Berger DCS. These makes dpm-g4cpp a suitable prototype for GPU/FPGA hardware implementations since the main thread divergence factors have been eliminated form the original DPM simulation.

Build

As mentioned before, the data generation and simulation phases are well separated. Since the data generation part depends on Geant4, a relatively recent version (>= Geant4-10.7p01) of the Geant4 simulation toolkit needs to be available on the system. See the Geant4 Installation Guide for further instruction.

It is assumed in the followings, that the required version of the Geant4 toolkit is installed on the system and the corresponding Geant4Config.cmake CMake configuration file is located under the G4_INSTALL directory. Then building and installing G4HepEm (under the G4HepEm_INSTALL directory) can be done by simple:

$ git clone https://github.com/mnovak42/dpm-g4cpp.git    
$ cd dpm-g4cpp
$ mkdir build
$ cd build/
$ cmake ../ -DGeant4_DIR=G4_INSTALL
$ make

Example: the data generation phase

Then the generated dpm_GenerateData executable can be used to generate all the files that are needed for a given simulation. The following information is reported when executing with -h (help) flag:

$ ./dpm_GenerateData -h

===============================================================================================
 The dpm-g4cpp pre-init/data generation phase.

 Usage: dpm_GenerateData [OPTIONS]

 -d  --input-data-dir      (where the pre-generated data will be written)  - default: ./data
 -c  --configuration-index (one of the pre-defined configuration settings) - default: 0
 -h  --help

===============================================================================================

Note, that there are several pre-defined simulation configurations available for the fast settings of some reference configurations given in [1]. See the Configuration.hh and the Simulation/Geom.hh files for further details. The pre-defined simulation configuration with the index of 3 will be used in the following: this will be used to simulate the depth dose distribution in a Water-Titanium-Bone-Water geometry shown in Fig.15. of [1] (and below). The corresponding data files will be generated into a newly created local data_WaterTiBoneWater directory:

bash-3.2$ mkdir data_WaterTiBoneWater
bash-3.2$ ./dpm_GenerateData -c 3 -d data_WaterTiBoneWater  

Expand for the output!

=== The dpm-g4cpp data generation configuration:

    - output data directory : data_WaterTiBoneWater
    - configuration index   : 3

=== Pre-init/data generation configuration:
     - Min-Ekin          : 0.001  [MeV]
     - Max-Ekin          : 21  [MeV]
     - #Ekins for e-     : 128
     - #Ekins for gamma  : 1024
     - cut for e-        : 0.2  [MeV]
     - cut for gamma     : 0.05  [MeV]
     - MSC step param    :
            s-low    = 5  [mm]
            s-high   = 10  [mm]
            e-cross  = 12 [MeV]
     - pre-defined geom. : 3
     - materials         :
            [0] = G4_WATER
            [1] = G4_Ti
            [2] = G4_BONE_COMPACT_ICRU
...............................................


=== Data are generated now (don't care about the
    Geant4 warning: 'WARNING from G4PenelopeIonisationModel...')

WARNING from G4PenelopeIonisationModel Atomic de-excitation module is not instantiated, so there will not be any fluorescence/Auger emission. Please make sure this is intended

=== All data have been generated and will be written now
    into files under the directory: data_WaterTiBoneWater

Example: the simulation phase

After having all the required data generated, the dpm_Simulate executable can be used to perform the dose simulation in the previously selected configuration. The simulation configuration can be specified by the input arguments. Executing dmp_Simulate with the -h (help) flag provides the following information regarding the input arguments:

$./dpm_Simulate -h

=================================================================================================
 The dpm-g4cpp simulation phase.

 Usage: dpm_Simulate [OPTIONS]

-p  --primary-particle    (possible particle names: e-, gamma)            - default: e-
-e  --primary-energy      (in [MeV] units)                                - default: 0.1
-n  --number-of-histories (number of primary events to simulate)          - default: 1.0E+5
-d  --input-data-dir      (where the pre-generated data are located)      - default: ./data
-b  --voxel-size          (size of the voxel/box in [mm])                 - default: 1.0
-c  --configuration-index (one of the pre-defined configuration index)    - default: 0
-h  --help

=================================================================================================

Therefore, the in order to simulate 200 000 primary e- with initial energy of 15.0 [MeV] using the previously generated data located under the data_WaterTiBoneWater sub-directory with the previously used pre-defined configuration with an index of 3 can be done as:

./dpm_Simulate -n 200000 -e 15.0 -d data_WaterTiBoneWater -c 3

Expand for the output!

=== The dpm-g4cpp simulation confguration:

    - input data directory  : data_WaterTiBoneWater
    - primary particle      : e-
    - primary energy        : 15 [MeV]
    - number of histories   : 200000
    - geometry voxel size   : 1 [mm]
    - confoguration index   : 3

=== Pre-init/data generation configuration:
     - Min-Ekin          : 0.001  [MeV]
     - Max-Ekin          : 21  [MeV]
     - #Ekins for e-     : 128
     - #Ekins for gamma  : 1024
     - cut for e-        : 0.2  [MeV]
     - cut for gamma     : 0.05  [MeV]
     - MSC step param    :
            s-low    = 5  [mm]
            s-high   = 10  [mm]
            e-cross  = 12 [MeV]
     - pre-defined geom. : 3
     - materials         :
            [0] = G4_WATER
            [1] = G4_Ti
            [2] = G4_BONE_COMPACT_ICRU
...............................................


=== Start simulation of N = 200000 events ===

== Starting Primary = 0
== Starting Primary = 20000
== Starting Primary = 40000
== Starting Primary = 60000
== Starting Primary = 80000
== Starting Primary = 100000
== Starting Primary = 120000
== Starting Primary = 140000
== Starting Primary = 160000
== Starting Primary = 180000

=== End simulation of N = 200000 events ===

=== Energy deposit histogram is written to the file:  hist.sim

Example: the result

The resulted depth dose distribution is located in the hist.sim file. This is plotted in the figure below together with the corresponding original DPM result (digitised from Fig.15. of [1] and can be found under tests/WaterTiBoneWater_15MeV/)

References

[1] J Sempau, S. J. Wilderman, A. F. Bielajew, Phys. Med. Biol. 2000 Aug; 45(8):2263-91