Ballroom C
Purpose: To develop and validate an automated procedure for the dosimetric assessment of treatment plan deliveries for upper gastrointestinal cancer sites on an online adaptive MR-Linac system.
Methods: The dose-of-the-day (DOTD) procedure was developed using the scripting environment in RayStation treatment planning system with the data generated in Online Monaco for Unity MR-Linac workflow. The method was applied to 16 patients, receiving 1 or 5 treatment fractions. Key data including pre-treatment MR images, contours, dose cloud from the adapt-to-shape online (ATS) plan as well as MR images acquired during treatment delivery (BON) were brought into RayStation for post-processing. An automated scripting pipeline was developed to perform: a) intensity-based deformable image registration (DIR) for ATS-to-BON image sets, b) deformable contour mapping ATS-to-BON, c) transfer rigidly BON contours on ATS, d) derive dosimetric analytics, and e) generate pdf report. The DIR process was validated by expert reviewers.
Results: The DIR mapped contours on BON data, were found clinically acceptable, provided that original ATS contours were adequate. The dose cloud discrepancies between ATS and BON were within ±30 cGy for max point values and ±3 cGy for mean dose. For most structures, the difference between the ATS and BON clinical goal values were within ±50 cGy. The outliers were observed for both 1 and 5 fraction regiments and different structures, suggesting no clear dependence on the site or fractionation scheme.
Conclusion: The intensity-based DIR performed well for ATS-to-BON data and allowed for prompt and reliable contour propagation between intra-faction MR images. The DOTD process was found to be an important clinical tool as it provided a quantitative assessment of the treatment quality while the patient was still on the bed. The reported dosimetric intra-fraction differences suggested the need for dose accumulation processes and further plan adaptation based on per session clinical goals.
Not Applicable / None Entered.
Not Applicable / None Entered.