ESTRO 2024 - Abstract Book

S3635

Physics - Dose prediction, optimisation and applications of photon and electron planning

ESTRO 2024

Mairead Daly 1 , Eliana Vasquez Ozorio 2 , Ananya Choudhury 1,3 , Alan McWilliam 2,4 , Ganesh Radhakrishna 3 , Cynthia L Eccles 5,1 1 The University of Manchester, Division of Cancer Sciences, Manchester, United Kingdom. 2 The University of Manchester, Radiotherapy Related Research, Manchester, United Kingdom. 3 The Christie Hospitals NHS Foundation Trust, Clinical Oncology, Manchester, United Kingdom. 4 The Christie Hospitals NHS Foundation Trust, Medical Physics and Engineering, Manchester, United Kingdom. 5 The Christie Hospitals NHS Foundation Trust, Radiotherapy, Manchester, United Kingdom

Purpose/Objective:

Online adaptive radiotherapy using the integrated magnetic resonance linear accelerator (MR Linac) is an emerging method for delivery of stereotactic ablative body radiotherapy (SABR) in locally advanced pancreatic cancer. Deformable image registration (DIR) is a tool that can be used to estimate accumulated delivered dose over a course of radiotherapy. However, accuracy of DIR strategies has not been deeply investigated for this application and it is thought to be of limited use within the abdomen. The purpose of this work was to evaluate the geometrical performance and its impact on estimating delivered dose for organs at risk (OAR) in pancreatic SABR of two DIR strategies: global and organ-wise (OW) DIR using magnetic resonance imaging (MRI).

Material/Methods:

Locally advanced, inoperable, non-metastatic pancreatic cancer patients at our institution are treated with 4000 cGy in 5# using online adaptive SABR on the MR Linac part of the MOMENTUM international registry study (NCT04075305 (1). For this study, data collected included the reference plan (Ref Plan), motion compensated 3D VANE MRI scan (pMRI), and pre-treatment MRI (FxMRI). All patients were imaged and treated with an abdominal compression belt. All scans, structure sets and doses were imported in RayStation v11B. To estimate the accumulated dose, the large bowel, small bowel loops, stomach, duodenum, left and right kidney cortex OARs were recontoured on each FxMRI throughout the treated volume, i.e., beyond the 2cm PTV expansion used online. Rigid grey level frame-of reference registrations used to initialise DIRs were created between pMRI and each FxMRI and edited manually to optimise the fit. Each FxMRI was registered to pMRI via two strategies: 1) Global DIRs, where the user changed similarity metric parameters (mutual information, correlation coefficient) to achieve the best registration possible and 2) OW DIR, where each of the re-contoured OARs were used as control structures in one independent registration per organ, per scan using the scripting interface. For each DIR strategy, its geometrical performance was evaluated using mean distance to agreement (mDTA) and Hausdorff distance (HD) between contoured and deformably mapped OARs. Moreover, the total delivered dose for each patient was calculated by mapping the fraction dose using each DIR strategy. Differences between planned and accumulated doses for mean and D0.5cc estimates are reported. Descriptive statistics were used to analyse dose difference and geometrical performance results. Statistical tests were not carried out due to the small sample size.

Results:

Table 1 summarizes the geometrical evaluation of both DIR strategies. OW DIR demonstrated superior geometrical performance with more optimal contour similarity over the global strategy. Mean percentage difference between accumulated doses and Ref Plan doses are shown in Figure 1. Differences ranged from -5.3 to 9.1% and 0.4 to 28.6% for mean doses (Fig 1a) when accumulating doses using global and OW DIR, respectively. A similar pattern is observed for D0.5cc, where these ranged -1.5 to -12.4 % and -9.2 to 36 % for mean doses (Fig