S1847
Physics - Dose prediction/calculation, optimisation and applications for photon and electron planning
ESTRO 2026
Purpose/Objective: Automated radiotherapy planning can improve efficiency, plan quality, and reduce inter-observer variability. Two-step workflows combining dose prediction and dose mimicking have shown promising results [1,2]. The purpose of this study was to develop and evaluate end-to-end fully automated VMAT planning strategies for breast cancer radiotherapy. Material/Methods: Two automated planning (AP) approaches were developed. AP-Method1 used a machine learning– based dose prediction trained on 48 VMAT plans from a European center using a moderate hypofractionation regimen (40.05 Gy/15 fractions) [3]. AP-Method2 used an analytical formula for generating a dose prediction [4]. Both methods shared the same dose-mimicking step to generate DICOM RT plans. External validation was performed on 10 patients. For each case, two automated VMAT plans were generated for a VersaHD linac, imported into RayStation for final dose calculation, and quantitatively compared with corresponding manual clinical plans using established evaluation criteria [3,5]. One physicist and two radiation oncologists reviewed the automated plans, adjusted normalization as needed, assessed clinical acceptability, and indicated their preferred method. Five cases underwent patient-specific QA to assess deliverability. Results: Automated plan generation averaged 8 minutes per case. Both AP methods produced plans that were clinically comparable or superior to manual ones, with improved CTV/PTV coverage and enhanced cardiac sparing (Table1). After minor normalization adjustments (mean 0.8%, range 0.4–1.4%), clinical acceptability was achieved in 90% and 93.3% of plans for AP-Method1 and AP-Method2, respectively (Table2). In terms of plan preference, expert opinions were split: 43.3% for AP-Method1, 53.3% for AP- Method2. Interestingly, assessment of inter-expert agreement showed that the average clinical acceptability between expert pairs (1–2, 2–3, and 1–3) was 80% for AP-Method1 and 87% for AP-Method2, whereas plan preference agreement was limited to 33%. Patient QA results demonstrated excellent deliverability, with 3%/2 mm gamma pass rates of 97% (AP-Method1) and 93% (AP- Method2).
Conclusion: The results indicate that the HyperSight iCBCT mode provides improved HU accuracy and dosimetric agreement compared to the Standard mode. The observed HU consistency and high gamma passing rates demonstrate the potential of iCBCT images for accurate Radiotherapy planning and adaptive radiotherapy applications. Further investigation with a larger patient dataset is recommended to validate these findings in clinical practice. References: - F. Dusi, et al., Evaluation of the dosimetric accuracy of HyperSight CBCT for CBCT-only adaptative radiotherapy workflow, Physica Medica, 2025.- N. Nelson, et al., Feasibility of HyperSight CBCT for adaptive radiation therapy: A phantom benchmark study of dose calculation accuracy and delivery verification on halcyon, Medical Physics, 2024. - C. Wessels, et al., Technical note: Phantom-based evaluation of CBCT dose calculation accuracy for use in adaptive radiotherapy, Medical Physics, 2024. Keywords: HyperSight CBCT, Dosimetry Accuracy Digital Poster 1608 Automated VMAT Planning for breast cancer radiotherapy Pauline Maury 1 , Ana-Maria Gardareanu 2 , Pascal Fenoglietto 3 , Gorkem Gungor 4 , Madalina-Liana Costea 5 , Baris Ungun 6 , Remi Vauclin 6 , Elie Mengin 6 , Nikos Paragios 7 , Waissi Waisse 8 1 Radiation Oncology Department, nstitute Gustave Roussy, Villejuif, France. 2 Radiation Oncology Department, nstitute Coltea Clinical Hospital, Bucharest, Romania. 3 Radiation Oncology Department, Institut du Cancer de Montpellier, Montpellier, France. 4 Radiation Oncology Department, Acibadem MAA University School of Medicine, Istanbul, Turkey. 5 Clinical Affairs, TheraPanacea, Paris, France. 6 Physics, TheraPanacea, Paris, France. 7 CEO, TheraPanacea, Paris, France. 8 Radiation Oncology Department, Centre Leon Berard, Lyon, France
Made with FlippingBook - Share PDF online