Iterative reconstruction of whole accelerator phase spaces for Intraoperative Radiation Therapy (IORT) from measured dose data

Monte Carlo (MC) methods are a very powerful tool to compute dose in radiotherapy, as all effects that need to be considered, such as material inhomogeneities, back-scatter from large bones, and beam hardening can be properly modeled. Their most important caveat, however, besides computation time, is that they need a realistic and reliable description of the electron and/or photon beam that delivers the dose, and this is not usually available. In this respect, Monte Carlo (MC) methods have been an invaluable tool for realistic modeling of medical therapy accelerators, including electron linear accelerators (LINAC) used in Intra-operative Radiation Therapy (IORT). The purpose of this work is to obtain the radiation beam properties (or phase-space fro the beam or PHSP) at phantom surface based on a set of dose measurements, without the need for a detailed simulation of the accelerator head and/or applicator. An iterative reconstruction algorithm (EM-ML), commonly used in tomographic image reconstruction, has been employed to optimize iteratively all aspects of the PHSP, such as energy spectra, particle type and fluency, and angle of particle emission, required by the MC dose calculation code Dose Planning Method (DPM). Phase space files for IORT have been derived for different energies and applicator diameters, which yield dose in good agreement with the measurements.