S2367
Physics - Quality assurance and auditing
ESTRO 2026
Belgium. 3 Department of Radiation Oncology, Inselspital, Bern University Hospital and University of Bern, Bern, Switzerland. 4 Center for Proton therapy, Paul Scherrer Institute, Villigen, Switzerland Purpose/Objective: CBCT-based synthetic CTs (sCTs) are commonly used in adaptive photon radiotherapy[1] but remain limited in use for proton therapy(PT) due to PT’s higher sensitivity to HU/stopping-power errors. Recent progress in sCT generation has achieved adequate quality in many cases, however, consistency remains an issue, and quality is still sometimes insufficient for replanning, hindering their routine clinical use. Additionally, with no ground truth available at adaptation time, there is no simple quantitative standard to judge whether a sCT is suitable for replanning. We hypothesized that combining an image-similarity metric with a dose-related metric computed between two sCTs from the same CBCT yield acceptability outcomes consistent with CT-based assessments. Our aim is to develop and validate a quantitative rule as a QA-tool for sCT-based replanning in head and neck PT. Material/Methods: We analyzed 184 CT-CBCT pairs from 51 head-and- neck cancer patients. From each CBCT, sCTs were generated in Raystation by deformable registration of CT-to-CBCT(vCT), analytical correction(CBCTcorr), and with an in-house deep-learning model(sCT-DL)[2]. The QA workflow compared sCTs generated by one reference(sCTref) method to another verification(sCTtest) method (each of the three could serve as either). The decision rule combines an image test (e.g.,structural similarity index measure (SSIM), Dice(bones)) with a dose test (e.g., Mean absolute error (MAE)), gamma pass rate (GPR)) (fig1). For validation only (not needed in clinical use), same-day CTs registered to the corresponding CBCTs served as ground truth(GT). Plans were re-optimized in Raystation on sCTref and doses compared with GT-CT using GPR2%/1mm to label sCTs as acceptable ( ≥ 95% voxel pass-rate) or unacceptable. Thresholds for image and dose were tuned on 35 test patients (129-pairs) to maximize balanced accuracy; 16 patients (55-pairs) formed an independent test set.
tumor volume (GTV) was 3.71 cc (0.15–10.32). The median maximum doses to the facial, trigeminal, and cochlear nerves were 25.41 Gy (17.93–28), 18.81 Gy (0.8–26.96), and 15.77 Gy (3.52–26.13), respectively.After a median follow-up of 4.5 months (2–12), only one patient experienced deterioration in hearing status. Interestingly, this patient received a cochlear dose below the cohort mean (13.30 Gy). Four patients reported new-onset tinnitus after treatment, though without associated hearing decline. No facial or trigeminal neuropathies were observed. Radiologically, all lesions remained stable in size, and one case demonstrated partial regression. Conclusion: This study provides additional evidence supporting the safety of the dose levels delivered to the cochlea and cranial nerves during hypo-FSRT for unilateral acoustic neurinomas. The observed dosimetric parameters were associated with minimal early toxicity, suggesting that the evaluated dose ranges may be considered acceptable for these organs at interest. These findings contribute valuable reference data toward defining preliminary dose constraints in hypofractionated stereotactic radiotherapy. Larger prospective studies with longer follow-up are warranted to validate these results and refine dose–toxicity correlations for critical neural structures. References: 1. Kalogeridi MA, et al. Stereotactic radiosurgery and radiotherapy for acoustic neuromas. Vol. 43, Neurosurgical Review. Springer; 2020. p. 941–9.2. Chen Z, et al. Five-year outcomes following hypofractionated stereotactic radiotherapy delivered in five fractions for acoustic neuromas: the mean cochlear dose may impact hearing preservation. International Journal of Clinical Oncology. 2018 Aug 1;23(4):608–14. 3. Puataweepong P, et al. Long- term clinical outcomes of stereotactic radiosurgery and hypofractionated stereotactic radiotherapy using the CyberKnife® robotic radiosurgery system for vestibular schwannoma. Asia-Pacific Journal of Clinical Oncology. 2022 Oct 1;18(5):e247–54. Keywords: acoustic neurinoma, nerve toxicity, hypo- FSRT Mini-Oral 2814 A fast and reliable Quality Assurance tool for CBCT- based synthetic CTs for the head and neck adaptive proton therapy Arthur Galapon 1 , Andreas Smolders 2 , Dirk Wagenaar 1 , Antje Knopf 3 , Arturs Meijers 4 , Stefan Both 1 , Francesca Albertini 4 1 Department of Radiation Oncology, University Medical Center Groningen, Groningen, Netherlands. 2 MIRO Laboratory, UCLouvain, Woluwe-Saint-Lambert,
Results: The highest accuracy on the validation set was achieved with a combination of SSIM( ≥ 0.96) and DoseMAE( ≤ 0.24cGy) resulting in a balanced accuracy of 97.8% on the test-set, with 100% sensitivity to
Made with FlippingBook - Share PDF online