ESTRO 2025 - Abstract Book

S3101

Physics - Inter-fraction motion management and offline adaptive radiotherapy

ESTRO 2025

Conclusion: The accuracy of the implemented traffic-light protocol was high. The minority of LA-NSCLC patients needed adaption at the investigated time-points, suggesting our IMRT planning procedures to be robust towards anatomical changes.

Keywords: Lung cancer, adaptive protocol, accuracy

1250

Digital Poster Daily proton dose re-calculation on deep-learning corrected cone-beam computed tomography for prostate patients Casper D Vestergaard, Ludvig P Muren, Stine E Petersen, Ulrik V Elstrøm, Liliana Stolarczyk, Ole Nørrevang, Vicki T Taasti Danish Centre for Particle Therapy, Aarhus University Hospital, Aarhus, Denmark Purpose/Objective: For synthetic CT (sCT) generation from cone-beam CT (CBCT) to be of benefit for adaptive proton therapy, the image quality and the accuracy of proton dose calculation for the sCTs must be high and stable across all treatment fractions. In this study, we developed a 3D CycleCUT deep-learning network for sCT generation [1] for prostate cancer patients and evaluated its performance across the full treatment course of 39 fractions. Material/Methods: The deep-learning network was trained on CTs and CBCTs from 25 high-risk prostate cancer patients and evaluated on six separate patients – all 31 patients had been treated with proton therapy of the prostate and elective lymph nodes. Patients in the test set had a planning CT, 39 CBCTs, and at least one repeat CT used for replanning. The generated sCTs were compared to the fan-beam CTs in terms of i) CT number accuracy and stability within spherical regions-of-interest (ROIs) in the bladder and prostate, ii) proton range calculation accuracy through single-spot proton plans, and iii) dose trends for the target coverage over the treatment course (for a single patient). Results: The sCTs demonstrated image quality comparable to fan-beam CTs, while preserving the CBCT anatomy (Figure 1). The mean CT numbers on the sCTs were comparable to fan-beam CTs (Figure 2 top), e.g. for the ROI in the bladder, the mean CT numbers ranged from 7 HU to 35 HU for sCTs, and from 4 HU to 34 HU for fan-beam CTs, across all patients and fractions. The median proton range differences over the 39 fractions were close to zero for all six patients, the largest median proton range difference being 1.9 mm. The proton dose calculations showed an excellent target coverage (≥99.6%) for the high-dose target over all fractions (Figure 2 bottom). However, under dosage was seen in a few fractions for the low-dose target due to inter-fraction movement between the high-dose and low-dose targets where the prostate is prioritized in the online rigid registration.