Clinical validation of an analytical procedure for in vivo PET range verification in proton therapy

Proton therapy is a treatment modality of increasing interest in clinical radiation oncology mostly because its dose distribution allows high dose conformality and a reduced integral dose compared to conventional radiation therapy based on X-rays and electrons. Its advantages are centered on the characteristic proton Bragg peak: a sharp maximum in the linear energy transfer of the projectile, that, if rightly positioned, makes the effect of the irradiation on the tissue more localized, thus increasing therapy efficacy and reducing side effects. The potential benefits of such physical selectivity have led to multiple studies addressed to the evaluation of uncertainty sources in the radiation therapy treatment process, from the initial treatment planning to the real patient dose delivery, thus providing insights into ways in which new imaging strategies are necessary to monitor, correct and adapt for possible errors. An interesting method to assess the geometric accuracy of the planned treatment delivery and to ensure the high quality of the proton therapy is to use Positron Emission Tomography (PET), which takes advantage of the nuclear inelastic reactions between protons and nuclei in the tissue during irradiation, producing small amounts of short-lived β+-emitting isotopes detectable soon after the irradiation. Verification of the therapy can be achieved by comparing and then quantifying differences between the PET images and the yield of the positron emitters predicted on the basis of the treatment planning system. The purpose of this study has been to extend an existing model and to implement and evaluate it in a fast and flexible framework to predict locally such activity distributions taking directly the reference planning CT and planned dose as inputs. The results achieved in this patient study highlighted the potential of the implemented analytical model, to monitor the dose delivery, proposing as a powerful substitution method to the sensitive and time-consuming Monte Carlo (MC) approach for the calculation of expected activity distributions.