May 8, 2012
Calculation uncertainties undermine dose verification
Radiation treatment with charged hadron beams is steadily growing. There are presently 38 operational particle centres administering radiotherapy, 20 more planned or under construction, and many more being considered. According to PTCOG, more than 96,000 patients had been treated by the end of 2011. And at the recent ICTR-PHE conference in Geneva, the message was clearly that ion (or hadron) beam therapy is alive and well and facing a bright future.
Hadron beam therapy can be extremely complex and, therefore, may require highly complex verification in order to optimally benefit from its theoretical advantages over photon therapy. Such advantages include improved ballistic properties leading to tighter dose conformation to the target volume, with better sparing of healthy tissue. Hadrons heavier than protons also deliver enhanced biological effectiveness in the tumour.
For photon beam irradiation, the photon fluence transiting the body may be captured with an imaging panel, which may then be used to quantify dose discrepancies. However, nothing similar exists for hadron beams. The primary particles in a hadron beam are completely absorbed in the body, which precludes the use of electronic portal imaging.
A brilliant solution to verify dose distributions in hadron beams was formulated many years ago. The nuclear reactions inherent in hadron beams cause secondary particles to be released from their point of creation. Capturing these particles outside the body during irradiation may offer opportunities to establish correlations between the incident hadron fluence and its energy transfer in the body, and the kinetic characteristics of the emitted particles.
One such dose verification method under investigation involves nuclear transformation of the body's own tissue nuclei into positron-emitting instable nuclei. Soon after its creation, the positron released in the β+ decay annihilates with an electron, giving rise to annihilation photons that can be detected using a PET-like detector geometry.
Many research papers have been devoted to the correlation between reconstructed PET activity distributions and the absolute or relative hadron dose distributions, the latter of which may be used to verify hadron ranges in tissue. Treatment verification is achieved by comparing the measured spatial activity distribution with a simulated prediction.
A key method in this verification process is Monte Carlo simulation. These simulations are quite complex and rely on cross sections to transform the hadron fluences into activity distributions. Different Monte Carlo codes, however, employ cross sections taken from different experimental databases or theoretical models. The codes may also differ in the physics transport models of the primary hadrons and their secondaries. These differences may cause inherent uncertainties for hadron dose verification.
In an effort to quantify differences in hadron-generated activity distributions in tissue-like materials, it was decided within the European ENVISION collaboration to perform an extensive study of these differences. The study involved workers from the Maastro Clinic in Maastricht (Frank Verhaegen, Enrica Seravalli), the Heidelberg Ion Beam Therapy Centre (Katia Parodi and team), IMNC in Orsay (Irene Buvat and team), Delft University of Technology (Dennis Schaart) and accelerator manufacturer IBA, based in Louvain la Neuve (Frederic Stichelbaut).
Four Monte Carlo codes (FLUKA, PHITS, MCNPX, GATE) were compared in their capability to predict the generation of β+-emitting nuclei in proton interactions with carbon and oxygen. In a recent Physics in Medicine & Biology paper, the results of this study were reported for proton beams of up to 200 MeV impinging on water and PMMA phantoms (Phys. Med. Biol. 57 1659).
The study found that large differences in cross sections led to a large difference in the production rate and spatial distributions of β+-emitting nuclei in different Monte Carlo codes. Discrepancies of more than 100% in production rates (reactions/proton) were found between simulations using internal models (GATE, MCNPX and PHITS) and simulations relying on experimental cross section datasets (FLUKA).
While hadron range verification, one of the important applications of nuclear activation based imaging, may not be severely influenced by the cross section uncertainties (Phys. Med. Biol. 56 2687), the situation regarding absolute dose verification is much more critical. This is because in the latter application, the absolute number of emitted positrons matters. As long as these cross section uncertainties exist, results from simulation studies must be regarded with caution. The accuracy of simulated results may be particularly hard to assess from studies employing a single Monte Carlo code without experimental validation. Additional uncertainties arise from the lack of accurate knowledge of tissue compositions and physiological washout in real patients.
The study clearly points to the need to provide more reliable data in Monte Carlo codes: mostly cross sections, but to a lesser extent also transport models and even ionization potentials. The latter were found to vary significantly among codes for a well known material such as PMMA. Although all the codes in this study used the same value of the ionization potential for water, that too has quite a high uncertainty (Phys. Med. Biol. 54 N205), which is ironic since water is arguably the most studied material in the universe. Although not the subject of this study, similar caution is most likely needed for other hadron beams such as carbon ions, where more complex fragmentation reactions are responsible for the hadron-generated activity distributions.
The group therefore highly recommends that an effort is made to investigate fundamental interaction data in hadron beams (cross sections, ionization potentials), which may involve extensive measurement campaigns at laboratories equipped for these kind of demanding measurements. Several institutes with operational proton beams have used empirical and often facility-dependent solutions to better reproduce β+ activity integral yields measured in thick tissue-like targets.
Monte Carlo simulations provide an indispensable tool with which to compare predicted to measured activity distributions. More effort is also needed to improve the interaction models in those cases where they are not based on measured cross sections. As long as the reported uncertainties remain, hadron dose verification may be on insecure footing indeed.
About the author
Frank Verhaegen is professor and head of clinical physics research at the Maastro Clinic in Maastricht, the Netherlands