ESTRO 2021 Abstract Book

S357

ESTRO 2021

R. Canters 1 , V. Taasti 1 , D. Hattu 1 , J. Van Loon 1 , D. De Ruysscher 1 1 GROW School for Oncology, Maastricht University Medical Centre+, Department of Radiation Oncology (Maastro), Maastricht, The Netherlands Purpose or Objective Treatment planning for lung intensity modulated proton therapy (IMPT) is currently a laborious iterative process, taking up to a day. In this study we developed a fast automated optimization method (AOM) that iteratively adjusts objective weights and goals to achieve robust and clinically acceptable plans. Materials and Methods Our AOM, using the Raystation treatment planning system, starts with an initial optimization to determine the basic dose shape. The lung and heart optimization objectives are tightened in an iterative process by either increasing objective weights or lowering the objective dose (Figure 1). Subsequently, a similar iteration is performed for the esophagus. Finally, robustness is evaluated using a voxelwise robustness evaluation and, if necessary, target objectives are further tightened. To validate this automated robust optimization approach, we compared automatically generated plans for 20 lung patients with the standard clinical plan. The evaluation cohort consisted of 4 small cell lung cancer patients (30x1.5Gy), 8 non-small cell lung cancer (NCSLC) patients receiving concurrent radiochemotherapy (30x2Gy), and 8 NCSLC patients receiving sequential radiochemotherapy or radiotherapy only (25x2.4Gy). Plans were evaluated on the most relevant clinical goals for targets and organs at risk.

Figure 1: Flowchart showing the structure of the automated robust optimization process, containing an initial optimization, a lung/heart dose reduction step, an esophagus dose reduction step, and a robustness evaluation and improvement step. Results From 20 generated treatment plans, 15 were immediately clinically acceptable. 5 Plans contained maximum body doses that were slightly higher than the clinical limit, which could be corrected in in a single additional optimization loop. On average, the mean lung dose, and lung V5Gy decreased with respect to the standard plans, while body max dose and mean esophagus dose increased slightly (p<0.05)(Figure 2). Mean heart dose, target coverage and spinal cord max dose did not change significantly.