ESTRO 2020 Abstract book

S123 ESTRO 2020

adjustments was quantified and the relevance was discussed with an experienced dosimetrist. Results Only 4.2% of the automated plans in the validation set were rejected according to the automated plan approval for violation of one or more constraints (Table 1). All but one of the captured Auto-Plans could indeed be improved with manual adjustments. Of the automatically accepted plans, 13% had actually been adjusted in clinical practice. However, none of these adjustments were judged clinically relevant with no significant difference in rectum Dmean (0.55Gy ± 1.2Gy), anal sphincter Dmean (0.22Gy ± 0.8Gy) and target V95% (0.01% ± 0.5%).

Conclusion The developed EGSnrc MRL model for the 1.5 T Elekta Unity showed good agreement with measured data. It therefore shows the general possibility for investigation of more complex research questions and future secondary dose calculations.

Conclusion Plan approval can be automatically performed with a higher specificity for capturing plans that could be improved, than a manual check by a dosimetrist. We are currently clinically implementing this framework, such that automatically approved plans (expected to be around 85%) would not be manually checked at all, significantly improving efficiency. OC-0221 Clinical experience of automated SBRT planning with Constrained Hierarchical Optimization L. Hong 1 , Y. Zhou 1 , J. Yang 1 , J. Mechalakos 1 , M. Hunt 1 , J. Yang 1 , J. Yamada 1 , J. Deasy 1 , M. Zarepisheh 1 1 Memorial Sloan Kettering Cancer Center, Medical Physics, New York, USA Purpose or Objective To present our clinical experience with a fully automated approach to SBRT treatment planning using Expedited Constrained Hierarchical Optimization (ECHO) to improve plan efficiency and quality. Material and Methods From April 2017 to September 2019, 1492 patients underwent SBRT radiotherapy for paraspinal and other metastatic tumors with 1739 different ECHO produced plans. From October 2019, ECHO also produced plans for post-brachytherapy prostate patients receiving 25Gy in 5 fractions SBRT treatment. After contouring, a template plan using 9 IMRT fields was set up and sent to ECHO through an application program interface plug-in from the treatment planning system Eclipse®. The clinical criteria required by the institution to be always met were formulated as hard constraints (such as maximum dose, mean dose and dose-volume-histogram (DVH) constraints), were strictly enforced by the optimization. Other clinical criteria defined as “desired” (e.g., better PTV coverage, lower normal organs’ doses) were optimized as much as possible by solving sequential constrained optimization problems. A correction step incorporating leaf sequencing and scattering contributions into optimization, and smoothing fluence map of beams for delivery efficiency was in the final step of ECHO. Upon ECHO completion, the planner received an email indicating the plan was ready for review.

Proffered Papers: Proffered papers 12: Artificial Intelligence and automation

OC-0220 Treatment planning without user interaction: Automatic plan approval of prostate Auto-Plans K. Kiers 1 , E. Van der Bijl 1 , J. Trinks 1 , A. Hogaarts 1 , R. Van der Bel 1 , G. Wortel 1 , E. Damen 1 , J. Tomas 1 1 The Netherlands Cancer Institute, Department of Radiation Oncology, Amsterdam, The Netherlands Purpose or Objective While automated treatment planning improves planning efficiency, in practice it is hard for dosimetrists to judge whether a plan can still be improved, resulting in unnecessary further optimizations. Using a combination of patient specific, delineation-based predictions, population statistics and strict clinical criteria, plans that potentially can be improved can be flagged and plan approval can be automated. The purpose of this work is to investigate the impact of fully automated treatment plan verification of prostate Auto-Plans on plan quality and efficiency. Material and Methods 100 treatment plans of prostate cancer patients with seminal vesicle invasion, without a hip prosthesis or bowel loop near the target, with a prescription of 20x3Gy were included and split in a train, test and validation set (42/10/48). The plans were generated using Pinnacle 3 Auto-Planning and automatically scaled to a sufficient coverage (Auto-Plans). In clinical practice Auto-Plans are evaluated manually and possibly adjusted (Clinical plans). The rectum- and anal sphincter Dmean and V95% were predicted using a linear regression model based on the PTV-OAR overlap volume histogram at 0 and 10 mm expansion. The conformity of the 80% isodose line was predicted based on the volume of the PTV, and thresholds on the PTV inhomogeneity were set based on population statistics such that the highest 5% would be captured. These thresholds and predictions, with a margin of 2 standard deviations to take in account inaccuracies of the models, were added to the existing general clinical criteria as personalized constraints to capture plans that can be improved even under the strict clinical criteria. The Auto-Plans in the validation set were evaluated on PTV coverage and - inhomogeneity, and the general and personalized OAR constraints. Plans that did not meet all criteria would be considered rejected and vice versa. This was retrospectively compared to which plans were adjusted in clinical practice. The magnitude of the