ESTRO 2021 Abstract Book

S318

ESTRO 2021

OC-0421 Towards a new optimization algorithm for Arc Proton Therapy treatment planning S. Wuyckens 1 , G. Janssens 2 , E. Sterpin 1,3 , J. Lee 1 , K. Souris 1 1 UCLouvain, Molecular Imaging, Radiotherapy and Oncology (MIRO), Brussels, Belgium; 2 Ion Beam Applications SA, Research department, Louvain-La-Neuve, Belgium; 3 KULeuven, Oncology department, Leuven, Belgium Purpose or Objective As Volumetric Modulated Arc Therapy (VMAT) has become the most common type of photon therapy treatment, Arc Proton Therapy (ArcPT) is now the logical evolution of conventional proton therapy. ArcPT plans are composed of a long discrete series of beam angles, the angular control points. Increasing the number of beam incidences is expected to increase treatment conformity and robustness. However, it would also significantly increase the delivery time. As a large part of the treatment time is spent switching the beam energy, the challenge of ArcPT plan optimization lies in selecting as few energy layers as possible, ideally only one, for each beam direction. In this work, we designed a beamlet-based algorithm for the optimization of proton arc therapy that distributes a number of energy layers similar to conventional therapy over more gantry angles. Materials and Methods The treatment is modeled as an optimization problem. In this work, the formalism initially proposed by UCLA (Gu W. et al. 2020) has been adapted by formulating the objective function as a weighted sum of objectives, composed of the usual dose fidelity term, a group sparsity regularization term, and an indicator function. The specificity of our formulation lies in the last two terms. A L2,1-norm was chosen to shrink the weights of all energy layers at each iteration and deactivate the least useful layers, whereas the indicator function counteracts this shrinkage by reactivating one energy layer in unused control points. These combined terms favor the selection of a single energy layer per control point as intended. The method FISTArcPT was tested on a lung tumor case and evaluated on these criteria: objective function value, layer sparsity and DVH metrics. The gantry rotation was assumed clockwise and the arc spanned angles from 180° to 270° with steps of 1.5°. For each beam, a spot grid was initially generated with a lateral spacing of 5 mm and a layer spacing of 5 mm over the target volume dilated by 10 mm. Results As illustrated in Fig. 1. an arc plan with 98.4% of the beams delivering only one energy layer per control point was successfully obtained with good dosimetry results comparable with the ones obtained with an IMPT plan made up of 3 beams. The dosimetry results are reported in the DVH in Fig. 2, with a target D95 reaching 95.6% of the prescription, a mean dose of 0.4 Gy for the heart, and a mean dose of 5.5 Gy for the right lung for both the IMPT and FISTArcPT plans. However, ArcPT has a dosimetric benefit compared with the IMPT plan since it achieved a decrease in the high doses received by the lung.