ESTRO 2024 - Abstract Book

S4404

Physics - Intra-fraction motion management and real-time adaptive radiotherapy

ESTRO 2024

Teo Stanescu 1,2,3 , Oleksii Semeniuk 4 , Andrea Shessel 5 , Michael Velec 5 , Michael Yan 6 , Aruz Mesci 6 , John Kim 6 , Laura Dawson 6 , Ali Hosni 6 1 Princess Margaret Cancer Centre, Medical Physics, Toronto, Canada. 2 University of Toronto, Radiation Oncology, Toronto, Canada. 3 University of Toronto, Mechanical and Industrial Engineering, Toronto, Canada. 4 The Warren Alpert Medical School of Brown University, Radiation Oncology, Rhode Island, USA. 5 Princess Margaret Cancer Centre, RMP, Toronto, Canada. 6 Princess Margaret Cancer Centre, Radiation Oncology, Toronto, Canada

Purpose/Objective:

This study presents the development, validation, and clinical implementation of an automated tool for rapidly assessing the dosimetric impact of intra-fraction motion on online adaptive treatment planning and delivery using an MR-Linac system.

Material/Methods:

The tool was created within the RayStation scripting environment and tested on 25 patients with liver, pancreas, and oligometastases who received 1, 3, or 5 fractions on an MR-Linac system. The automated pipeline analyzed pre treatment MR images, contours, and dose data from the online adapt-to-shape (ATS) Monaco plan, along with MR images from the verification (VER) and treatment delivery beam-on (BON) MR scans. Key steps involved deformable image registration (DIR), mapping deformable regions of interest (ROI), dosimetric and similarity analyses for relevant ROIs, and report generation.

Results:

A retrospective analysis covered 84 treatment fractions. Two expert reviewers assessed over 1000 anatomical structures to validate the DIR process. For 24 ATS plans recomputed on VER and BON data in Offline Monaco, the maximum discrepancy in the dose clouds was within 1% of the prescription dose. For most luminal structures, the ATS-VER and ATS-BON dosimetric differences in clinical goal values were within ±100 cGy per fraction for the maximum point dose. The analysis also detected larger outliers not apparent during the default treatment workflow, unrelated to fractionation regimen or anatomical structure. ROI similarity analysis between ATS and VER/BON within the treatment region strongly correlated with intra-fraction changes in OARs, flagging dosimetric discrepancies.

Conclusion:

The dose assessment tool is crucial for the online adaptive clinical workflow, offering an on-demand mechanism to validate online-generated treatment plans before delivery. Dosimetric differences from BON data collected during treatment can help develop strategies, such as dose accumulation processes, to adjust clinical goals before subsequent treatment fractions.

Keywords: Online adaptive MR-guided RT, dose verification