ESTRO 2024 - Abstract Book

S70 ESTRO 2024 The acquisition of a computed tomography (CT) image using X-rays before the start of treatment is classically required for dose calculation in the current radiotherapy (RT) workflow. Invited Speaker

This imaging modality exposes the patient to ionizing radiation, introduces uncertainties in magnetic resonance (MR) to CT image registration, and is an additional process in the RT workflow.

In an MR-only RT workflow, artificial intelligence is used to generate synthetic CT (sCT) images from MR data.

The ultimate goal is to eliminate the need for CT imaging and its associated disadvantages.

This presentation explores the implementation of an MR-only workflow, but also shares insights to overcome challenges related to imaging, patient positioning, and clinical implementation.

Standardization of MR sequence imaging protocols is fundamental, considering the discrepancy in acquisition times between the two imaging modalities, to minimize patient motion.

Optimization of the patient setup involves placing the coils over the immobilization devices and introducing new fiducial markers to be able to define the user of origin and the couch position during planning.

Furthermore, some limitations were observed in the workflow and translated into exclusion criteria, which prevent such patients from undergoing the novel workflow.

These patients are then treated according to the conventional workflow, which, requiring both MR and CT imaging, poses challenges in scheduling the extra simulation, identifying the exclusion criteria and decision-making.

The MR-only RT workflow has been successfully clinically implemented for non-stereotactic brain cases and male pelvis cases, with ongoing efforts for other anatomical regions.

3437

Multi-criteria optimisation with AI aspects for brachytherapy treatment planning

Björn Morén

Linköping University, Department of Mathematics, Linköping, Sweden

Abstract:

High dose-rate brachytherapy is used to treat a number of sites, for example prostate, cervix, and head and neck, each with their specific treatment protocol and characteristics. The focus of this talk is the inverse treatment planning problem. The two main parts of inverse treatment planning are the catheter placement and applicator selection, and the dwell time optimization. A variety of evaluation criteria are used to evaluate the resulting dose distribution, for example DVH parameters, visual inspection, properties of the dwell time distribution and TCP/NTCP. The treatment aims are in conflict, most notably local control and sparing of organs at risk, and the necessary trade-offs are patient specific. Hence, it is natural to consider a multi-objective formulation of the inverse treatment planning to be able to take well-informed decisions for each patient. A short history of treatment planning models and the development will be given, starting with single-objective models based on penalty functions, including parameter tuning to handle the multiple criteria, towards the current focus on