ESTRO 2022 - Abstract Book

S786

Abstract book

ESTRO 2022

that perFRACTION LFO has a better performance in detecting geometric errors as follows: in 52% of cases for 3%/3 mm/Global Th5%, in 57% of cases for 3%/2 mm/Global Th10%, in 61% of cases for 2%/2 mm/Global Th5%, in 42 % of cases for 1.5%/1.5 mm/Global Th10%, in 42% of cases for 1.5%/1.5 mm/Global Th5% and in 42% of cases for 1%/1 mm/Global Th10%. Of notice, Matrixx is always better than perFRACTION LFO in detecting MU -1% -2% dose error scenarios. Considering a 5% detection level perFRACTION LFO needs a gamma criterion equal or lower than 1.5%/1.5 mm (global) in order to detect geometric errors.

Fig.1 Passing rates for different error scenarios and different gamma criteria.

Conclusion Our error scenario analysis clearly shows that the perFRACTION LFO strategy has the potential to catch clinically relevant geometric errors if an adeguate gamma criterion is used. Combining LFO with transmission dosimetry methods (eg. EPID based) could increase the ability of the system in detecting clinically relevant dose delivery errors.

PD-0894 atlas-based treatment planning models for magnetic resonance guided therapy

A. Khalifa 1,11 , J. Winter 2,3 , I. Navarro 4,4 , C. McIntosh 5,6,4,7,8,9 , T.G. Purdie 5,10,4,3

1 University of Toronto, Department of Medical Biophsyics, Toronto, Canada; 2 Princess Margaret Cancer Centre, Radiation Medicine Program, Toronto, Canada; 3 University of Toronto, Department of Radiation Oncology, Toronto, Canada; 4 Princess Margaret Cancer Center, Radiation Medicine Program, Toronto, Canada; 5 University of Toronto, Department of Medical Biophysics, Toronto, Canada; 6 Techna Institute, University Health Network, Toronto, Canada; 7 University Health Network, Peter Munk Cardiac Center, Toronto, Canada; 8 University Health Network, Joint Department of Medical Imaging, Toronto, Canada; 9 Vector Institute, Vector Institute, Toronto, Canada; 10 University Health Network, Techna Institute, Toronto, Canada; 11 University Health Network, Techna Institute , Toronto, Canada Purpose or Objective Atlas based machine learning (ML) for radiation treatment planning effectively creates new treatment plans by selecting the most anatomically similar patients (i.e., atlases) from a training database and mapping their dose distributions to a novel patient’s unique anatomy. We investigated whether an atlas-based ML model, trained only on computed tomography (CT) imaging, could generate clinically applicable treatment plans in the MR guided setting by directly predicting dose on MR imaging without model retraining. Materials and Methods We included contoured CT and 3D T2 MR imaging from prostate cancer patients (n=55) treated on a 1.5T MR-linac. For each patient, we generated moderately hypofractionated (60 Gy in 20 fractions) VMAT treatment plans on both images using an atlas-based ML treatment planning model, trained on CT imaging only. Statistically significant differences between dose distributions of MR- and CT-based treatment plans were identified using DVH metrics, as per institutional evaluation criteria, and Wilcoxon Signed Rank tests ( α = 0.05). To determine if these dosimetric differences were due to anatomical differences between images, they were compared to differences in relative PTV overlap of the rectum and bladder walls between images, calculated as the volume of OAR-PTV intersection divided by the OAR volume. To determine if changing the input image influenced model predictions (i.e., by selecting dosimetrically differing atlases in anatomically similar cases), differences in overlap and DVH metrics between patients and their selected atlases were compared for each image type. Results Statistically significant changes in dose-volume metrics between MR- and CT-based plans were identified for PTV D99, bladder wall D30 and D50, and rectum wall D50 (Figure 1). Differences in the amount of PTV overlap between the MR and CT images moderately correlated (r > 0.7, p < 0.001) with differences in dose-volume metrics (Figure 2a). Differences in overlap between a patient and their selected atlases correlated (r > 0.7, p < 0.001) with the dose difference between them (Figure 2b), and this relationship was consistent between CT- and MR- based plans