ESTRO 2023 - Abstract Book

S1344

Digital Posters

ESTRO 2023

The most complicated anatomical regions to segment were the larynx, bladder, and rectum, with a DSC of 0.41±0.04, 0.48±0.40, and 0.50±0.20, respectively, and HD of 23±8 mm, 65±85 mm, and 37±29 mm. The AB and DL models generated CTV_LN contours in 8.4±0.6 min, and OARs in 4.3±0.2 min. Since Sept. 2022, all TMLI patients at our Institute are pre segmented using DL and AB models; the RO reviews and manually adjust the CTV_LN and OARs, with a time-saving >50% with respect to previous manual contouring procedure. Conclusion DL and AB models can streamline the RT process for TMLI treatments. A hybrid approach, where the human operator reviews and modifies structures after automatic generation remains fundamental, in particular for CTV_LN delineation.

This work was supported by grant GR-2019-12370739 .

PO-1646 KB plan optimization models’ transferability: multi-institutional international consortia validation

L. Placidi 1 , R. Castriconi 2 , G. Benecchi 3 , M. Burns 4 , E. Cagni 5 , P. Griffin 6 , C. Markham 7 , K. McGoldrick 8 , V. Landoni 9 , E. Moretti 10 , C. Oliviero 11 , V. Panettieri 12 , G. Rambaldi Guidasci 13 , T. Rancati 14 , A. Turner 15 , A. Scaggion 16 , A. Tudda 17 , C. Fiorino 17 1 Fondazione Policlinico Universitario A. Gemelli IRCCS, UOSD Medical Physics and Radioprotection, Roma, Italy; 2 IRCCS San Raffaele Scientific Institute, Medical Physics Department , Milano, Italy; 3 University Hospital of Parma AOUP, Medical Physics Department, Parma, Italy; 4 Peter MacCallum Cancer Centre, Box Hill Campus, Melbourne, Australia; 5 Azienda USL IRCCS, Department of Advanced Technology, Reggio Emilia, Italy; 6 The Alfred Hospital, Alfred Health Radiation Oncology, Melbourne, Italy; 7 Peter MacCallum Cancer Centre, Parkville Campus, Australia, Australia; 8 Peter MacCallum Cancer Centre, Parkville Campus, Melbourne, Australia; 9 IRCSS Regina Elena National Cancer Institute, Department of Medical Physics, Rome, Italy; 10 Udine University Hospital, Department of Medical Physics, Udine, Italy; 11 University Hospital, ‘‘Federico II”, Medical Physics, Napoli, Italy; 12 The Alfred Hospital, Alfred Health Medical Physics, Melbourne, Australia; 13 Fatebenefratelli Isola Tiberina – Gemelli Isola, UOC di Radioterapia Oncologica, Roma, Italy; 14 Fondazione IRCCS Istituto Nazionale dei Tumori, Prosate Cancer Program, Milano, Italy; 15 Peter MacCallum Cancer Centre, Bendigo Campus, Bendigo, Australia; 16 Veneto Institute of Oncology IOV–IRCCS, Medical Physics Department, Padova, Italy; 17 IRCCS San Raffaele Scientific Institute, Medical Physics Department, Milano, Italy Purpose or Objective The Knowledge-based (KB) approach may efficiently translate planning experiences into models able to classify patients according to their individual anatomical features and possibly to automatically drive optimization on new patients. The reliability of model predictions between institutes may be due to poor overlap between the geometrical/anatomical features of Institutional training cohorts. The aim of current investigation was to assess models’ transferability between two multi-institutional international consortia (one in Europe and one outside Europe) in the case of tangential field (TF) irradiation of right whole breast (RWB). Materials and Methods In the context of a national multi-institutional project (consortium A), ten institutes set their KB models using RapidPlan (Varian Inc.) for RWB irradiation using TF. Intra-consortium transferability was evaluated by the overlap of the geometric Principal Component (PC1) when applied to the test patients (two from each institution) of the other 9 institutes. This parameter was assessed as representative of the association between anatomy/geometry and DVH prediction of OARs, quantifying if a model can be reasonably applied to a patient owning to another Institute. With the same methodology, twenty patients belonging to another national multi-institutional consortium (consortium B) were used to perform cross inter-consortium validation tests. Each model of consortium A was evaluated on 20 patients provided by consortium B. The degree of transferability of all models was quantified by the number of cases in which the ipsilateral lung PC1 was within the 90th percentile of the training set. Results Figure 1 shows transferability values in terms of the percentage of patients with the ipsilateral lung PC1 within the 90th percentile of the training set for both consortia. For consortium A, test results already demonstrated a high inter-institute transferability, with a PC1 value inside the 90th in 9 out of 10 models for more than 90% of the patients.