ESTRO 2021 Abstract Book

S1579

ESTRO 2021

1 Hospital de la Santa Creu i Sant Pau, Medical Physics, Barcelona, Spain

Purpose or Objective Lung cancer Radiotherapy treatment planning is the paradigm of dose calculation accuracy in the presence of low-density heterogeneities. Therefore, the use of a dose calculation engine that could accurately account for them such as Monte Carlo or Deterministic linear Boltzmann transport equation (D-LBTE) solvers would be desirable. However, clinical endpoints are based on dose calculation engines based on the convolution of point or pencil kernels that can only account partially for the lateral charged particle transport. We aimed to identify those images feature descriptors that could inform us when D-LBTE solver leads to large dose differences with respect to standard algorithms. Materials and Methods From 2012, SBRT lung cancer patients at our institution underwent a short-range 4DCT scan followed by a CT scan with 3 mm slice thickness, both under moderate abdominal compression. The former was used to delineate a MIP-based ITV, while the latter was used to calculate a 3DCRT plan in Eclipse using from 8 to 11 coplanar/non-coplanar 6 MV fields. AAA versions 8.9, 13.5 and 15.6 were used with a calculation grid of 0.25 cm. In this retrospective study, we used Eclipse Automation to search for 229 plans in the Aria database and to recalculate them all using both the same version (15.6) of AAA and AXB algorithms and the same calculation grid of 0.125 cm with fixed MU. For each patient ITV and PTV volume (TV), density (HU) histogram of ITV and the ring between ITV and PTV boundaries (PTV-ring) were gathered as predicting factors. The differences between algorithms were assessed using the variation of the target volume covered by the prescription isodose (TV PIV ) and the Paddick conformity index (CI Paddick , ICRU 91). We carried out two types of analysis. The first was to correlate the patients’ descriptors to the differences in target volume coverage and the Paddick conformity index. The second was to find any breakpoints (BP) using the method described in Muggeo VM (2003). Results Figures 1 and 2 present the differences of PTV volume covered by dose prescription (ΔTV PIV ) and Paddick conformity index between AAA and AXB in terms of PTV-ring median density. We found breakpoints with statistical significance (p<0.01) in - 738±6 HU and -783±5 HU. The difference between the analysed algorithm increased considerably for patients with PTV-ring densities lower than the BP. Both parameters had weak correlation with volume (-0.259 and 0.186). When considering the median density of the ITV and the PTV-ring we found a lower dispersion for PTV-ring variable.

Figure 1: Coverage difference in PTV volume covered by dose presciption (AAA-AXB)