The elemental concentration (especially oxygen and carbon) and mass density must be accurately assigned to perform Monte Carlo (MC) simulations for predicting proton-induced nuclear reactions in the human body. We recently proposed an approach to quantify elemental concentrations and mass densities of human soft tissues from water content (WC) data obtained by quantitative magnetic resonance (MR) imaging (which we called "MRWC"). This study presents the first implementation of MRWC-derived elemental concentrations and mass densities as complementary inputs into MC simulations on a virtual head phantom, and demonstrates the simulation of positron emitter production yields in proton therapy. An MC code, PHITS, was used to simulate proton therapy with a monoenergetic 140 MeV beam for a digital head phantom provided by BrainWeb. Three different head images were synthesized as inputs: a conventional CT image, an ideal CT image as a reference, and a WC image coupled with the bone-only CT image for a hybrid approach (MRWC/CT). Thereafter, the performance of the MRWC/CT method was evaluated by comparing its accuracy in predicting the production yields of positron emitters (11C and 15O) with the gold-standard CT-only method. The MRWC/CT method could predict 11C and 15O production yield maps that closely resembled the corresponding reference maps, while the CT-only method failed. The structural similarity index measures between the reference and CT- or MRWC/CT-derived maps were improved from 0.67 (CT-only) to 0.87 (MRWC/CT) for 11C and 0.76 (CT-only) to 0.93 (MRWC/CT) for 15O. Furthermore, applying post-processing normalizations to account for elemental density variations in the production yields of positron emitters facilitated the determination of distal fall-off positions in depth activity profiles. At least in the head area, the MRWC/CT method demonstrated potential for more precise predictions of proton-induced positron emitter distributions via MC simulations than that of the CT-only method.
Read full abstract