ESTRO 2022 - Abstract Book

S164

Abstract book

ESTRO 2022

A.Mascia

USA Abstract not available

Symposium: Automatic planning: Towards a sustainable future

SP-0204 Keeping automatic planning up to date: How to incorporate changes to dose prescription, technique and OAR constraints

L. Marrazzo 1

1 Careggi University Hospital, Medical Physics Unit, Florence, Italy

Abstract Text Commonly used interactive (‘manual’) inverse plan optimization for IMRT (Intensity Modulated Radiation Therapy) and VMAT (Volumetric Modulated Arc Therapy) is workload and time intensive, while plan quality may heavily depend on planner’s expertise. In the last few years, automatic planning (AP) has been introduced with the aim of reducing planning time and inter- operator variability and improve plan quality. Different approaches to automated planning can be used, the most common methods being knowledge-based planning (KBP), protocol-based automatic iterative optimization (PB-AIO), and multicriteria optimization (MCO) which can be pareto-navigation driven or automated. Some kind of solution is now available in most commercial treatment planning systems. The process of introducing automatic planning in clinical practice is generally laborious, requiring algorithm configuration and validation. KBP configuration includes creation of OAR DVH prediction models, based on a library of high-quality previous plans, PB-AIO requires the creation of an optimization template where the right parameters (clinical requests and other planning parameters) must be set, and MCO needs a careful definition of planning constraints and objectives functions, which has to be supplemented with objectives priorities and goal values in case of wish-list driven automated MCO. Since a suboptimal configuration translates into a plan quality systematically lower than feasible, a validation of any AP approach prior to clinical introduction is mandatory, comparing automatically generated plans with high-quality manually generated plans. Typically, AP configurations are for specific delivery approaches, clinical sites, clinical protocols (doses to targets and OARs) and TPS versions, and the actual manual planning quality is used as a starting point. But radiation therapy is a rapidly evolving field, where new prescription strategies, new fractionation schemes, new volume definitions, new dose constraints and new delivery approaches are regularly introduced. Moreover, changes in treatment planning systems, e.g. implementation of new optimizers, new cost functions or new configurations may give rise to opportunities for enhancing quality of automatically generated plans. Also advances in manual plan quality may be a stimulus to further enhance AP plan quality. In this presentation we will discuss continuing efforts to keep automated planning up to date.

SP-0205 Automated planning for online adaptive radiotherapy

M. Palacios

The Netherlands Abstract not available

SP-0206 How can treatment planning be imrpoved if nobody can do standard planning anymore?

L. Rossi 1

1 The Netherlands Abstract not available

SP-0207 Autoplanning for brachytherapy: Does it work?

R. Bijman 1

1 Erasmus Medical Center, Radiotherapy, Rotterdam, The Netherlands

Abstract Text After all the promising results and years of experience with automated treatment planning in EBRT including several different techniques such as IMRT, VMAT, IMPT, SBRT, etc. it is a logical step to explore automated treatment planning for brachytherapy (BT). Several implementations have been proposed in the literature (i.e ISPA, HIPO, GOMEA), all with their benefits and drawbacks. At our institute we developed software for automated multi-criterial optimization for EBRT fluence map optimization, Erasmus-iCycle. With Erasmus-iCycle, fully automated and high quality treatment plans were generated for a large variety of EBRT treatment sites and modalities. Recently its functionality was extended with BT dwell time optimization (BiCycle) for a set of fixed dwell positions distributed over the intracavitary, ovoids and needles. The software is wish-list driven and capable of optimizing simultaneously dosimetrical and geometrical (i.e. dwell time distribution) aims. The system has been validated in the past for prostate cancer and cervical cancer. The promising results showed that