ESTRO 2020 Abstract book

S162 ESTRO 2020

PD-0291 Comparison of different autosegmentation software for left-sided breast cancer patients J. Heikkilä 1 , T. Viren 1 , H. Virsunen 1 , K. Vuolukka 1 , L. Voutilainen 1 , R. Sawabi 2 , H. Abouelazm 2 , M. Kauppila 3 , J. Baeza 4 , W. Van Elmpt 4 , S.U. Akram 5,6 , J. Korhonen 3 , J. Seppälä 1 1 Kuopio University Hospital, Center of Oncology, Kuopio, Finland ; 2 University of Eastern Finland, Department of Applied Physics, Kuopio, Finland ; 3 Kymenlaakso Central Hospital- Kymenlaakso Social and Health Services Kymsote, Medical Imaging and Radiation Therapy, Kotka, Finland ; 4 GROW - School for Oncology and Developmental Biology- Maastricht University Medical Centre, Department of Radiation Oncology MAASTRO, Maastricht, The Netherlands ; 5 Aalto University, Department of Computer Science, Helsinki, Finland ; 6 MVision AI, MVision, Helsinki, Finland Purpose or Objective Autosegmentation of organs at risk (OAR) and target volumes (CTV) can be used to reduce contouring time and decrease inter- and intra-operator variation in the contouring task. The purpose of this study is to compare the accuracy of different autosegmentation software for delineation of CTVs and OARs of left-sided breast cancer patients. Material and Methods Breast CTV and OARs where manually contoured for 25 breast cancer patients according to the ESTRO guideline by expert radiation oncologists. Subsequently, patients were divided in to atlas/training set (n=20) and the test set (n=5). An artificial intelligence (AI) and four atlas- based segmentation software were used: MVision (Mvision AI Oy, Helsinki, Finland), ABAS (Elekta AB, Stockholm, Sweden), MIM (MIM Software Inc, OH, USA), Eclipse (Varian Medical Systems Inc, PA, USA) and Mirada (Mirada Medical Ltd, Oxford, UK). The contours created to five test patients were evaluated by calculating six different metrics to quantify the similarity between the automatically and manually delineated volumes: Dice similarity coefficient (DSC), Jaccard index (Jac), volume similarity (VS), average distance (AVE), BF score (BF) and Hausdorff distance (HD). Results CTV Breast, Heart, Left Lung and left anterior descending artery (LAD) structures were selected for the analysis. The autosegmented contours were compared to the manual delineation. An example of delineations for one test patient is shown in Figure 1. Evaluation metrics were calculated for all five test patients and average values (± SD) of all these parameters are shown in Table 1.

delineation.

Table 1: Average (± SD) values of evaluation metrics (DSC, Jac, VS, BF, HD and Ave) of autosegmented CTV Breast, Heart, Left Lung and LAD structures for five test patients. Five different software was used for autosegmentation (ABAS, Mirada, MIM, Eclipse and MVision).

Conclusion Performance of autosegmentation software was similar for CTV breast, lung and heart structures. In terms of Ave metric, all software segmented CTV structures were within 5 mm compared to manual reference delineation. More variation was detected in the autosegmentation of LAD between the software. Even though the number of patients in the training set was limited, AI method resulted equal or better contours as compared to atlas based methods.

PD-0292 Breast cancer patients with brain metastases: A multicenter retrospective study (KROG 16-12)

Abstract withdrawn

PD-0293 The benefit of high hyperthermia dose in re- irradiation of loco-regional recurrent breast cancer. A. Bakker 1 , J. Van der Zee 2 , G. Van Tienhoven 1 , H.P. Kok 1 , C.R.N. Rasch 3 , H. Crezee 1 1 Amsterdam UMC, Radiation Oncology, Amsterdam, Netherlands Antilles ; 2 Erasmuc MC, Radiation Oncology, Rotterdam, Netherlands Antilles ; 3 LUMC, Radiation Oncology, Leiden, Netherlands Antilles Purpose or Objective Hyperthermia therapy (HT), heating tumors to 40-45°C, is a known radiotherapy and chemotherapy sensitizer. The additional benefit of HT to re-irradiation (re-RT) for loco- regional recurrent breast cancer has been proven in multiple randomized trials. However, the efficacy of re- RT+HT varies widely. We performed a systematic review to investigate whether there is a relationship between achieved HT dose, tumor response and thermal toxicity for patients with recurrent breast cancer treated with re- RT+HT. Material and Methods Four databases, EMBASE, PubMed, Cochrane library and clinicaltrials.gov, were searched with the terms breast, radiotherapy, hyperthermia therapy and their synonyms. Final search was performed on April 3 rd 2019. The search was not limited to any date, both articles and conference proceedings were allowed, a filter for English articles and

Figure 1: An example of autosegmented structures for one patient. The filled structure is manual referense