Introduction Limits of Radiotherapy Treatment Planning System (TPS) when computing dose in moderate density media are known. Previous studies were published using Monte Carlo methods [1] . When treating lung cancer during Deep Inspiration Breath Hold (DIBH), lung density can decrease up to 0.12 g cm−3. The goal of this study is to assess performances of a TPS (Eclipse AAA) and of a MC model under these particular conditions: small fields and very low densities. Films measurements in heterogeneous phantoms are compared to both computation systems. Material and methods MC computation and model of the TrueBeam Linac head (Varian, Palo Alto, CA) are achieved using GATE/Geant4. Geometrical dimensions as well as the space phase files are furnished by Varian [2] . Jaws, mylar and baseplate are modeled and validated for the following energies: X6, X10, X6FFF and X10FFF. For this validation, MC computation are compared with Semiflex (PTW) ion chamber measurements in a water tank for a large range of field sizes (3 × 3 to 20 × 20 cm2). Percentage depth dose (PDD) and lateral dose profiles at different depths obtained by measurement and computation are compared (Gamma Index method). Two different qualities of cork (densities of 0.24 and 0.12 g cm3 are used to build two heterogeneous water/lung phantoms. These densities correspond to the lung densities during free breathing and DIBH respectively. Phantoms are composed of 20 × 20 cm2 slabs: PMMA (3 cm) + cork (7 cm) + PMMA (15 cm). See Fig. 1. Dose is measured at several depths using EBT3 gafchromic films (Ashland). Field sizes were 3 × 3 and 8 × 8 cm2 for X6FFF and X10FFF photon fields. These phantoms are implemented both in Eclipse (AAA) and MC. Download : Download high-res image (121KB) Download : Download full-size image Results TrueBeam MC model and measurements are in good agreement for all field sizes in both water and heterogeneous phantoms. More than 99% of PDD and lateral profiles passed the gamma index test (2% 2 mm). In lung heterogeneity, differences are observed between AAA and measurement results. These differences were higher for small field and very low density. As an example, for a 3 × 3 cm2 field at density 0.12 g cm−3, dose at a depth of 7 cm was 56.61%, 56.25%, and 67.54% for measure, MC and AAA respectively. Conclusion For very low densities, AAA shows its limits when MC is in good agreement with measurements. This MC model will allow us to assess our TPS results in complex cases like lung patients treated by Stereotactic Radiotherapy during DIBH.