ESTRO 2024 - Abstract Book

S357

Beachytherapy - Physics

ESTRO 2024

An applicator has been designed for the treatment of superficial lesions of rectal cancer. The results obtained in the measurements performed on the prototype are considered optimal for the future clinical use of the applicator.

Work will continue on the development of a new prototype in the same material to improve the fixation of the applicator on the patient in the operating theatre.

Keywords: IORT, rectal cancer, applicator prototype

1681

Poster Discussion

Intuitively customizable AI-based cervix brachytherapy treatment planning

Leah R Dickhoff 1 , Ellen M Kerkhof 1 , Heloisa H Deuzeman 1 , Laura A Velema 1 , Danique L Barten 2 , Bradley B Pieters 2,3 , Carien L Creutzberg 1 , Peter A Bosman 4 , Tanja Alderliesten 1 1 Leiden University Medical Center, Radiation Oncology, Leiden, Netherlands. 2 Amsterdam UMC University of Amsterdam, Radiation Oncology, Amsterdam, Netherlands. 3 Cancer Center Amsterdam, Cancer Treatment and Quality of Life, Amsterdam, Netherlands. 4 Centrum Wiskunde & Informatica, Evolutionary Intelligence, Amsterdam, Netherlands

Purpose/Objective:

BRIGHT is an AI-based approach to automated BT treatment planning. Unique to BRIGHT are its ability to directly optimize DVH aims [1] and the generation of a set of high-quality treatment plans with different trade-offs between target coverage and OAR sparing that enables insightful final plan selection. BRIGHT is currently clinically used for prostate cancer [2]. Besides common planning aims, such as the EMBRACE-II planning aims for cervix BT, there are often additional aims stemming from local clinical preferences. Here, BRIGHT is extended to intuitively support local preferences through a customizable set of aims as an additional objective. We showcase its use for cervix BT and its flexibility by employing four different customizations and evaluating the clinical quality of resulting plans.

Material/Methods:

Using three objectives is readily supported by the AI-based optimization algorithm in BRIGHT. Two objectives concern target coverage and OAR sparing aims according to the EMBRACE-II protocol. A third objective concerns customizable added aims (see Table). We defined these for our own clinic in an iterative feedback loop with a multi-disciplinary clinical team. BRIGHT then searches for optimal trade-off plans by actively focusing on the currently least-satisfied aim in each objective. The algorithm is halted after 3.7 min (NVIDIA RTX A6000 GPU), which was found sufficient to achieve convergence. Optimization is done using 220,000 dose calculation points (DCPs). Final plans are re-evaluated using 500,000 DCPs. 1/3 to each other (where V cc/point is the volume one DCP represents), while receiving a minimum dose. Considering contiguous volumes during optimization is a capability distinctive to BRIGHT, since it does not require gradients. We are particularly interested in contiguous volumes with a minimum dose of 250% and size of at least 0.125 cm 3 . Using a hard A contiguous volume can be defined as follows: all DCPs which are closer than 2(V cc/point )