Author: Amy Berrington de Gonzalez, Keith T Griffin, Jae Won Jung, Cari M Kitahara, Choonik Lee, Choonsik Lee, Aimee L McNamara, Matthew Mille, Harald Paganetti, Sergio Morato Rafet, Jan PO Schuemann, Jungwook Shin, Torunn I Yock 👨🔬
Affiliation: University of Michigan, East Carolina University, Massachusetts General Hospital, National Cancer Institute, Division of Cancer Epidemiology and Genetics, National Cancer Institute, Massachusetts General Hospital and Harvard Medical School, Institute of Cancer Research 🌍
Purpose: Radiotherapy dose reconstruction enables retrospective epidemiological studies that connect late health effects to the organ-absorbed radiation doses received during treatment. Though clinical dose distributions are available from the treatment planning system (TPS), these doses are typically partial-body and are known to be inaccurate for anatomical regions with heterogeneous density and for regions far out-of-field. Monte Carlo (MC) simulations are considered the gold-standard method for addressing such issues but require significant computational resources. The aim of this study is to highlight the dosimetric benefits of running MC simulations for organ-absorbed dose reconstruction within proton therapy patients.
Methods: Patient data were retrieved for fifty pediatric patients treated with passive scattering proton therapy at Massachusetts General Hospital. An in-house, automated workflow was developed on a supercomputing environment to reconstruct the patient-specific treatment within the TOPAS MC package and produce dose and dose-volume estimates throughout the body. Comparisons were made against results available from the TPS pencil beam algorithm.
Results: The average MC dose reconstruction consumed 12,000 CPU-hours. Dose differences were exhibited within near-field normal tissues (e.g., esophagus, brain substructures) of up to 2 Gy, with greater dose differences in regions of heterogeneous tissue density. An overall trend of increasing percent difference was seen between TPS and MC results with decreasing absolute dose, since the TPS algorithm does not compute dose at distances beyond a few centimeters from the treatment field. The TPS does not report neutron dose, which was found to be on the order of 1-100 mGy, and is considered to be more carcinogenic than proton dose.
Conclusion: Epidemiological studies that rely on MC simulations can expect some substantial improvements in near-field and out-of-field dose reconstruction with a relatively high tradeoff in computational cost. MC tools also provide the capability to investigate the biological effectiveness of low to moderate neutron doses.