ESTRO 2024 - Abstract Book

S3614

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

ESTRO 2024

Lorenzo Placidi 1 , Roberta Castriconi 2 , Peter Griffin 3 , Giovanna Benecchi 4 , Mark Burns 5 , Elisabetta Cagni 6 , Cathy Markham 5 , Valeria Landoni 7 , Eugenia Moretti 8 , Caterina Oliviero 9 , Giulia Rambaldi Guidasci 10 , Guenda Meffe 1 , Tiziana Rancati 11 , Alessandro Scaggion 12 , Alessia Tudda 2 , Vanessa Panettieri 13 , Karen McGoldrick 14 , Claudio Fiorino 2 1 Fondazione Policlinico Universitario Agostino Gemelli IRCCS, Medical Physics Unit, Roma, Italy. 2 IRCCS San Raffaele Scientific Institute, Medical Physics Dept, Milano, Italy. 3 The Alfred Hospital, Alfred Health Radiation Oncology, Melbourne, Australia. 4 University Hospital of Parma AOUP, Medical Physics Dept, Parma, Italy. 5 Peter MacCallum Cancer Centre, Department of Radiation Therapy Services, Melbourne, Australia. 6 Azienda USL-IRCCS, Department of Advanced Technology, Medical Physics Unit, Reggio Emilia, Italy. 7 IRCSS Regina Elena National Cancer Institute, Department of Medical Physics, Roma, Italy. 8 University Hospital, Department of Medical Physics, Udine, Italy. 9 University Hospital ‘‘Federico II”, Medical Physics Unit, Napoli, Italy. 10 Fatebenefratelli Isola Tiberina, UOC di Radioterapia Oncologica, Roma, Italy. 11 Fondazione IRCCS Istituto Nazionale dei Tumori, Data Science Unit, Milano, Italy. 12 Veneto Institute of Oncology IOV–IRCCS, Medical Physics Department, Padova, Italy. 13 Peter MacCallum Cancer Centre, Department of Physical Sciences, Melbourne, Australia. 14 Peter MacCallum Cancer Centre, Department of Radiation Therapy Services, Moorabbin, Australia

Purpose/Objective:

Multi-institutional knowledge-based (KB) plan prediction models have previously been generated to assess differences in plan performance between a national consortium (Consortium A) of institutions in the context of right whole breast (RWB) irradiation using tangential fields (TF). The current investigation aimed to quantify the resulting KB plan predictions in an multi-institutional consortium (Consortium B), quantifying their “geographical” transferability.

Material/Methods:

Ten institutions belonging to the multi-institutional Consortium A generated RWB-TF KB models using RapidPlan (Varian Inc.). The models were previously trained and validated using a common methodology, such as number of patients (> 70), contouring (national guidelines for CTV/PTV/OARs), techniques (wedged or FiF) and outlier elimination criteria, which were jointly discussed and defined [1]. Inter-institute interchangeability was assessed within Consortium A based on ipsilateral lung principal component (PC1) values calculated on an internal cross validation cohort (20 patients). Similarly, 20 patients from another national automated planning consortium (Consortium B) were used to perform an external cross-validation data test. The transferability of the models between consortia and the assessment of the dosimetric model predictions were performed. Each model from Consortium A (ten models) was tested on 20 patients provided by Consortium B. The transferability of the models was determined by the number of cases in which the ipsilateral lung PC1 was between the 10th and 90th percentile of the training set. Dosimetric parameters were also analysed. The ipsilateral lung mean dose and V20Gy were calculated to quantify the variability of the plan prediction between consortia and compared with the clinical DVH. In addition, the prediction of ipsilateral lung DVH was quantified and grouped as follows: optimal prediction (cDVH is within the predicted DVH band), suboptimal prediction (cDVH is below the predicted DVH band), improved prediction (clinical DVH is above the predicted DVH band) and failed prediction (predicted DVH band strongly disagrees with the cDVH).

Results: