The GEANT4 toolkit for microdosimetry calculations: application to microbeam radiation therapy (MRT).

Theoretical dose distributions for microbeam radiation therapy (MRT) are computed in this paper using the GEANT4 Monte Carlo (MC) simulation toolkit. MRT is an innovative experimental radiotherapy technique carried out using an array of parallel microbeams of synchrotron-wiggler-generated x rays. Although the biological mechanisms underlying the effects of microbeams are still largely unknown, the effectiveness of MRT can be traced back to the natural ability of normal tissues to rapidly repair small damages to the vasculature, and on the lack of a similar healing process in tumoral tissues. Contrary to conventional therapy, in which each beam is at least several millimeters wide, the narrowness of the microbeams allows a rapid regeneration of the blood vessels along the beams' trajectories. For this reason the calculation of the "valley" dose is of crucial importance and the correct use of MC codes for such purposes must be understood. GEANT4 offers, in addition to the standard libraries, a specialized package specifically designed to deal with electromagnetic interactions of particles with matter for energies down to 250 eV. This package implements two different approaches for electron and photon transport, one based on evaluated data libraries, the other adopting analytical models. These features are exploited to cross-check theoretical computations for MRT. The lateral and depth dose profiles are studied for the irradiation of a 20 cm diameter, 20 cm long cylindrical phantom, with cylindrical sources of different size and energy. Microbeam arrays are simulated with the aid of superposition algorithms, and the ratios of peak-to-valley doses are computed for typical cases used in preclinical assays. Dose profiles obtained using the GEANT4 evaluated data libraries and analytical models are compared with simulation results previously obtained using the PENELOPE code. The results show that dose profiles computed with GEANT4's analytical model are almost indistinguishable from those obtained with the PENELOPE code, but some noticeable differences appear when the evaluated data libraries are used.