ESTRO 2024 - Abstract Book

S3684

Physics - Dose prediction, optimisation and applications of photon and electron planning

ESTRO 2024

Material/Methods:

Our in-house LRT optimizer has been developed using MATLAB 2022a (Mathwroks). LatticeOpt uses as input the patient’s CT, structure set and the value of sphere diameters (user-defined) to generate the optimal Lattice structure set in a DICOM-RT format. First of all, the automatic tool finds the 3D-spatial configurations that maximise the number of vertices within the GTV. As a second step, it chooses the solution that minimise the overlapping volume histogram (OVH) [10] of the user-selected Organs at Risk (OARs), or, if no OARs are selected, it minimises the Hausdorff distance between vertices and GTV boundary, to reach a uniform vertex distribution. To validate the LatticeOpt, CT and structure set data from a cohort of 10 patients enrolled in a clinical trial between 2008 and 2016 were retrospectively included in this study. The median (range) of the clinical target volume is 901 cc (165; 3274). All lattice structures were generated according to a previously validated work [5]. LatticeOpt was evaluated in terms of number of vertices and OARs dose sparing. To validate the OVH optimization, for each patient, two lattice vertex arrangements were created, with and without OVH minimization. Secondly, the corresponding two plans were generated employing 6MV photons delivered with Volumetric Modulated Arc Therapy technique from a C-arc LINAC. The same optimization function template was used for both plans in the Treatment Planning System (TPS, Eclipse Varian v.13.7), so that the different results depend only on the lattice spatial configuration. Plan differences were evaluated in terms of mean dose and maximum dose to the selected OARs, as well as in terms of vertex dose coverage (D98%) and valley-vertices dose difference, (gradient index score) [11] within the GTV. Statistical analyses were performed using Wilcoxon rank test.

Results:

For each patient, LatticeOpt automatically found the maximal possible number of vertices within the GTV. The treatment plans based on the two lattice arrangements (OVHopt, OVHunopt) have similar lattice distribution with no statistical difference in vertex coverage and gradient index score. Successful OAR sparing was observed in all 10 patients (Figure 1) with a 4% and 9% difference for mean and max dose, respectively (p-values < 0.01 for both cases).