ESTRO 2021 Abstract Book

S1436

ESTRO 2021

We have created a simple incident learning system (ISL) for our department internal use using available Microsoft Office 365 features. The ISL is expected to help us collect and accumulate relevant, accurate and reliable data to perform consistent and continuous Failure Modes and Effects Analysis FMEA at our department and compare our results to published data and reports. PO-1709 Automation of DVH constraint checks and physics quality control improves patient safety N. Jensen 1 , I. Wahlstedt 1 1 Rigshospitalet, Oncology, Copenhagen, Denmark Purpose or Objective This study investigates whether patient safety can be enhanced by a quick implementation of a scripting- based automated electronic checklist (PlanCheck) for physics quality control review (QCR). Materials and Methods PlanCheck runs on a plan-by-plan basis using the Eclipse Scripting Application Programming Interface and evaluates both technical aspects and site-specific dose volume histogram (DVH) constraints for all radiotherapy photon plans at our institution. To determine whether the automated checks of technical aspects of the plan improves patient safety compared to manual QCR, 331 consecutively approved radiotherapy plans previously reviewed with manual QCR, were retrospectively checked with PlanCheck. In order to quantify the additional impact of the automated DVH constraint checks on patient safety, 47 consecutively approved breast cancer plans, previously reviewed with manual dose constraint check, were retrospectively checked with PlanCheck. Results 433 (3.4%) of the 12783 automated technical checks executed in the 331 plans yielded an error. All errors were scored using the severity rating from the AAPM TG-100 report. 19 of these errors (4%) either could have affected or affected the target dose (severity 5+) implicating a maximum dose difference to the target or a critical organ at risk of 0.5% to 10% and 3 errors could have resulted in stereotactic brain treatments being delivered to the wrong location of the brain (severity 10). In the 47 breast cancer plans, retrospectively subjected to automated DVH check, 10 undocumented dose constraint violations were found, varying between 0.1 Gy and 14.5 Gy above clinical constraint. Conclusion We have shown that automating the physics QCR using a method demanding minimum time and programing skills improves patient safety compared to manual QCR by experienced medical physicists. PO-1710 Atlas-based auto-segmentation for delineating the heart and cardiac substructures radiation therapy M.L. Milo 1 , T.B. Nyeng 2 , E.L. Lorenzen 3 , L. Hoffmann 2 , D.S. Møller 2 , B.V. Offersen 4 1 Aarhus University Hospital, Department of Experimental Clinical Oncology, Aarhus, Denmark; 2 Aarhus University Hospital, Department of Medical Physics, Aarhus, Denmark; 3 Odense University Hospital, Laboratory of Radiation Physics, Odense, Denmark; 4 Aarhus University Hospital, Department of Experimental Clinical Hospital, Aarhus, Denmark Purpose or Objective In early breast cancer (BC) radiation therapy (RT), a dose-response relationship between mean heart dose (MHD) and the risk of cardiac toxicity has been demonstrated. However, the dose distribution in the heart is heterogeneous with the highest dose deposition in the ventral part, and there are conflicting results regarding MHD as a predictor for cardiac toxicity. To support studies investigating radiation-associated cardiac toxicity, automatic segmentation is desirable in terms of reproducibility and efficiency. The aim of this study was to develop and validate an automatic atlas-based method for delineating the heart and cardiac substructures in non-contrast enhanced planning CT scans of BC patients. Materials and Methods The atlas database consisted of non-contrast enhanced planning CT scans from 42 BC patients, each with manual delineation of the heart and cardiac substructures. The auto-segmentation was developed in the MIM software system and validated geometrically and dosimetrically in two steps: A first validation in CT scans of six BC patients and a final test in a dataset of CT scans of 12 BC patients. For geometric evaluation, two metrics were used: The Dice Similarity Coefficient (DSC) and the Mean Surface Distance (MSD). For dosimetric evaluation, the RT doses to the heart and cardiac substructures in the manual and the automatic delineations were compared. Results The first validation showed high overlap between the auto-segmentation and manual delineation with DSC at 0.94 for the heart, and DSC ranging between 0.81-0.94 for the four cardiac chambers. A linear correlation between RT doses for the automatic and manual delineations was observed for the heart and cardiac substructures. The final test confirmed a high agreement between the automatic and manual delineations for the heart (DSC 0.94) and the cardiac chambers (DSC 0.76-0.88). DSC decreased with smaller structures. The MSD for the automatic delineations was slightly higher than the MSD for the manual delineations. The difference in MSD between the automatic and manual delineations was lowest for the larger structures and increased with smaller structures. In general, the difference in MSD between the automatic and manual delineations was low (<3mm) in virtually all structures. Finally, a high correlation between mean RT doses for the automatic and the mean RT doses for the manual delineations was observed for the heart and cardiac substructures. Conclusion An automatic CT-based atlas for delineation of the heart and cardiac substructures in BC RT was developed and validated with acceptable deviations between the automatic and manual delineations. The atlas will be