Monte Carlo dose calculation in presence of low-density media: Application to lung SBRT treated during DIBH

Purpose: Commercial algorithms used in Radiotherapy include approximations that are generally acceptable. However their limits can be seen when confronted with small fields and low-density media. These conditions exist during the treatment of lung cancers with Stereotactic Body Radiation Therapy (SBRT) achieved with the "Deep Inspiration Breath Hold" (DIBH) technique. A Monte Carlo (MC) model of a linear accelerator was used to assess the performance of two algorithms (Varian Acuros and AAA) in these conditions. This model is validated using phantoms with different densities. Lastly, results for SBRT cases are compared to both Acuros and AAA. Methods: A Varian TrueBeam linac was modeled using GATE/Geant4 and validated by comparing dose distributions for simple fields to measurements in water and in heterogeneous phantoms composed of PMMA and two types of cork (corresponding to lung densities during free-breathing and DIBH). Experimental measurements are also compared to AAA and Acuros. Finally, results of Acuros/AAA are compared to MC for a clinical case (SBRT during DIBH). Results: Based on 1D gamma index comparisons with measurements in water, the TrueBeam model was validated (>97% of points passed this test). In heterogeneous phantoms, and in particular for small field sizes, very low density (0.12 g.cm(-3)) and at the edge of the field, MC model was still in good agreement with measurements whilst AAA and Acuros showed discrepancies. With the patient CT, similar differences between MC and AAA/Acuros were observed for static fields but disappeared using an SBRT arc field. Conclusions: Our MC model is validated and limits of commercial algorithms are shown in very low densities. (C) 2017 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.