ESTRO 2022 - Abstract Book

S570

Abstract book

ESTRO 2022

Conclusion The combination of machine learning based dose prediction and robust dose mimicking optimization can be used to automatically create clinically acceptable robust IMPT plans for OCP non-inferior to manual treatment plans.

MO-0638 Novel optimization functions designed for re-irradiation treatment planning

J. Ödén 1 , K. Eriksson 1 , S. Svensson 1 , E. Setterquist 2 , J. Lilley 3 , C. Thompson 3 , C. Pagett 3 , A. Appelt 4 , L. Murray 5 , R. Bokrantz 1

1 RaySearch Laboratories AB, Research, Stockholm, Sweden; 2 RaySearch Laboratories AB, Development, Stockholm, Sweden; 3 Leeds Cancer Centre, St. James' University Hospital, Medical Physics and Engineering, Leeds, United Kingdom; 4 Leeds Cancer Centre, St. James' University Hospital, Medical Physics and Engineering, Leeds Institute of Medical Research at St James's, University of Leeds, Leeds, United Kingdom; 5 Leeds Cancer Centre, St. James' University Hospital, Clinical Oncology, Leeds Institute of Medical Research at St James's, University of Leeds, Leeds, United Kingdom Purpose or Objective Re-irradiation (reRT) for locoregional recurrence or second cancers after previous radiotherapy (RT) is a promising treatment technique, though anatomical changes between courses and radiobiologically-appropriate dose summation of multiple RT courses are recognised challenges. Optimally, reRT treatment planning should take the previously delivered dose into account. However, if traditional normal tissue threshold dose levels are already exceeded in previous treatments, regular optimization functions might cause suboptimal reRT plans. We propose a set of novel optimization functions designed for reRT treatment planning. Materials and Methods The standard physical optimization functions in RayStation that quadratically penalize voxel doses above a threshold dose level were adjusted to quadratically penalize the total equivalent dose in 2 Gy fractions (EQD2) above a threshold EQD2 level (EQD2 level ). The EQD2 level was further adjusted in a voxelwise fashion to account for the accumulated delivered EQD2 and a selected minimum allowed reRT EQD2 (EQD2 delivered+reRTmin ) as, where EQD2* level [ x ] is the adapted EQD2 level for voxel x in a ROI. An α / β =3 Gy was used for all voxels in this study. To account for partial tissue recovery between the RT courses, a set of reRT scale factors ([0,1]) that scale the EQD2 for each previous RT course could be selected. These adjustments were implemented in a research version of RayStation 11A for the maximum EQD2/EUD/DVH functions and the EQD2 fall-off function (Fig. 1). To demonstrate proof of concept, reRT VMAT plans of 35 Gy in 5 fractions (EQD2=70 Gy) were optimized to fulfil D 95% ≥ 95% and D 2% ≤ 105% using standard and novel optimization functions (EQD2 reRTmin =1–2 Gy) for a spherical relapse (20 cc) in three positions (scenarios A, B and C) following a fictive pelvic treatment of 60 Gy in 20 fractions (EQD2=72 Gy) on a human phantom (CIRS 801-P) (Fig. 2). The homogeneity index (HI=EQD2 95% /EQD2 5% ), EQD2 conformity index (CI=V target, 95% /V 95% ) and EQD2 gradient index (GI=[V 20% -V 80% ]/V target ) were evaluated. EQD2* level [ x ] = max(EQD2 level [ x ], EQD2 delivered+reRTmin [ x ]),