There have been rapid developments in recent years in using tools based on Monte Carlo (MC) simulation for medical physics applications. The challenges of more accurate estimations have led to the need for customised simulation tools and the use of larger computers. MC simulations are essential for characterizing the machines and sources used in proton and ion therapy and evaluating new designs while they are also important for nuclear medicine and medical imaging. They can also be used for better treatment planning but also for retrospective studies as they help to decide about the dose and the duration of the therapy. Finally, they prove to be extremely powerful in simulating regions with very different materials or with a high density gradient, as can be the case in some human body organs (i.e. lung, brain) but also in cases of complicated tissue/bone interfaces.

A large number of applications of radiation transport tools have existed for the past decades in medical physics. Recently, new tools, such as Geant4 have been adopted fervently by medical physicists to design improved diagnostic tools and to model the radiation dose delivered to patients in proton and hadron therapy. The number of journal publications with Geant4 applications for medicine is growing each year.

Geant4 is the latest in a series of software packages for High Energy Physics (HEP) studies in which physicists have to simulate large number of particles and how they interact with matter. It has been developed by an international collaboration of which CERN is a leading member. Geant4 is capable of simulating all particle types including those important for medicine: gammas, electrons, positrons, protons, neutrons, and ions. It includes models for all important physics interactions: electromagnetic, hadronic, photolepton-hadron and optical. Many of the models are of direct relevance to both the High energy physics and medical physics domains.

Development of Geant4 started in 1994 and resulted in a first release in 1998. It was quickly picked up by scientists in medical physics and today the Geant4 “toolkit” paper is one of the most cited publications, according to Thomson-Reuters' "Nuclear Science and Technology" and "Instruments and Instrumentation". This serves to underline the wide use of the package in many fields, in particular those outside particle physics. Geant4 has many applications in a wide variety of topics from PET, SPECT and CT scanners, external-beam design etc.

One of Geant4’s most prominent features for medical applications is its ability to handle complex geometries, as it was designed from the outset to handle the incredibly complex geometries of the detectors of the LHC experiments. Geant4, offers the most flexible geometry modeller of any particle and radiation transport tool. This important feature is fully exploited when dealing with complex medical equipment, including accelerators, collimators and shielding, as well as with complex models of patients derived from CT imaging.

Simulations with Geant4 are used to describe simple brachytherapy sources, multi-leaf collimators through to the most complex parts of proton IMRT machines. Moreover, Geant4 is used to model sources and geometries that have moving parts, also known as 4D Monte Carlo. This is crucial as many human-body organs are in constant motion; the space occupied by our lungs constantly changes as we breathe, and all other major organs move as a result. Also a 4D tool allows simulation in cases of imaging with moving scanners, dynamic multi-leaf collimators or rotating parts of proton IMRT machines.

Another important feature of Geant4 is that it is the first particle transport tool coded in C++, a modern Object Oriented programming language. This has allowed Geant4 to be extended easily with additional geometry shapes and specialised physics models from many different authors, and this has been achieved without creating an overhead in its computational performance.

Today, different communities of medical physicists and software engineers are working to exploit the capabilities of GEANT4 for medical applications through the development of specialized packages. Perhaps one of the most popular is GATE (GEANT4 Application for Tomographic Emission) that supports simulations of Positron Emission Tomography (PET), Single Photon Emission Computed Tomography (SPECT), Computed Tomography (CT) as well as the simulation of radiotherapy experiments. GATE is used extensively to simulate and predict scanner performance; one of its first jobs was to support the development of the ClearPET small-animal PET scanner that was developed by the CERN-CMS group. Other packages built using Geant4 are GAMOS, G4DNA, G4EMU, G4NAMU, TOPAS and PTSim.

The validation of GEANT4 by medical physicists and Geant4 collaborators requires a major effort that has been done in collaboration with members of the SFT group, who focus on validating physics models for use in LHC and other HEP experiments. Intensive use of Geant4 by medical communities provides extra information about Geant4 accuracy and identified potential issues relevant to HEP applications. Issues identified during validation have resulted in important improvements in the physics models, which have made them more accurate and stable for use in HEP and other domains. For example, validation of the multiple scattering model for ion chambers provided the stimulus for an improvement which has improved the accuracy of the energy deposition in sampling calorimeters, including those of LHCb and ATLAS. Furthermore, recent improvements of the multiple scattering models in Geant4 has been driven by the need for separate models for e, μ and hadrons, in order to obtain better agreement with the enlarged set of electron scattering data available for comparison. Refined models are now in production use in the LHC experiments.

A difficult challenge that had to be faced has been the simulation of ion interactions, given the many processes that have to be taken into account as they interact with nuclear matter. Significant work was done to integrate “standard” HEP-oriented and “low-energy” medical-and-space oriented electromagnetic-process packages. This allows the fine-grained choice of model for each application, independent of its domain. Refined models for bremsstrahlung and the photo-electric processes borrowed heavily from the low-energy implementation and the low energy part of proton and ion ionisation models is obtained from the original implementation in the low-energy package.

Another field of current development concerns efficient navigation in complex geometries. The regular structure of voxel phantoms and other geometries imported from CT scans can be described using an extension called ‘nested parameterisation’ that allows to vary the material of a volume depending on its location. An alternative treatment uses a 3-D voxelization directly and allows a particle to cross volume boundaries within a single simulation step, if all the voxels contain the same material. In the most recent development, a new solution has been developed to describe the emerging polygon-based geometries, which define a human phantom in terms of organs defined as polygon meshes. An external library, DAGMC, which provides efficient navigation in the most complex polygon meshes, was interfaced to the Geant4 solid interface.

At the recent conference “Geant4 2013 International User Conference on Medicine and Biology applications”, (Bordeaux, France, 7-9 October 2013) new results on validation of Geant4 electromagnetic and hadronic interactions were presented. There were several presentations related to the description and usage of very detailed geometry, and for the analysis of electromagnetic, hadronic, and ion interactions with matter. The OpenGATE collaboration reported on their method of using GPU devices for significant speedup of simulation. So, technology transfer from HEP to medicine in return receive new validation results of medical users and new ideas useful for future development of Geant4 for HEP.

You can visit the SFT webpage for the Geant4 project here. 

The author would like to thank: John Apostolakis, John Harvey and Vladimir Ivantchenko for their help and their thoughtfull comments.