S1843
Physics - Dose prediction/calculation, optimisation and applications for photon and electron planning
ESTRO 2026
Leon Berard, Lyon, France. 3 Radiation Oncology department, Institut du Cancer de Montpellier, Montpellier, France. 4 Radiation Oncology department, Acibadem MAA University School of Medicine, Istanbul, Turkey. 5 Radiation Oncology department, Coltea Clinical Hospital, Bucharest, Romania. 6 Clinical Affairs, TheraPanacea, Paris, France. 7 Physics, TheraPanacea, Paris, France. 8 CEO, TheraPanacea, Paris, France Purpose/Objective: Automation in radiotherapy treatment planning is a rapidly evolving field aiming at increasing planning efficiency, improving plan quality, and reducing inter- observer variability. Promising results have been reported using a two-step auto-planning (AP) workflow combining knowledge-based dose prediction and dose mimicking [1,2]. The purpose of this study was to develop and evaluate two end-to-end automated VMAT planning methodologies for head and neck (HN) radiotherapy in oropharyngeal cancer patients. Material/Methods: Two AP methodologies were implemented. For the first method (AP-Method1), a machine learning-based dose prediction model was trained using datasets from two large European centers. For the second method (AP-Method2), a dose prediction was generated using a published analytical formula [3]. Aside from the dose prediction approach, both methods used an identical dose mimicking component. External validation was performed on 8 patients from one center. For each case, two automated VMAT treatment plans were generated for a VersaHD linac to deliver 70Gy to the high-risk and 54.25Gy to the low-risk target in 35 fractions. The RTplan output was imported into Monaco TPS for final dose calculation. The resulting doses were qualitatively and quantitatively compared against corresponding manually generated clinical plans following guideline-based evaluation criteria [4]. Two medical physicists and two radiation oncologists reviewed the automated plans, adjusted the plan normalization if necessary, and assessed clinical acceptability. They also indicated preference between the two automated solutions. All plans underwent patient-specific QA measurements to verify deliverability. Results: The average total generation time per automated plan was approximately 10 minutes. Both AP methods produced clinically comparable or superior results relative to manual plans, showing improved PTV coverage and enhanced parotid sparing (Table1). After minor normalization adjustments (mean 2%, range 1.4–2.6%), all automated plans were deemed clinically acceptable. The overall clinical acceptability rate was 81.3% for AP-Method1 and 93.8% for AP-Method2. Two experts also reviewed the manual clinical plans,
Conclusion: The correlations between the mean heart dose, the Vx dose parameter, and the mean dose to the LAD and the left ventricle suggest that the mean heart dose alone is not the most important parameter for predicting heart toxicity. With the VMAT, the reduction in dose to the LAD and left ventricle suggests a potential reduction in heart toxicity, despite similar mean heart doses. References: [1] Darby et al. Risk of ischemic heart disease in women after radiotherapy for breast cancer. N Engl J Med 2013;368:987-998. Keywords: Breast treatment, heart dose Mini-Oral 1538 Automated VMAT Planning for Oropharynx Head and Neck Radiotherapy Frederic Gassa 1 , Anais Chardon 2 , Pascal Fenoglietto 3 , Gorkem Gungor 4 , Ana-Maria Gardareanu 5 , Madalina- Liana Costea 6 , Baris Ungun 7 , Remi Vauclin 7 , Elie Mengin 7 , Nikos Paragios 8 , Marie-Claude Biston 1 1 Medical Physics department, Centre Leon Berard, Lyon, France. 2 Radiation Oncology department, Centre
Made with FlippingBook - Share PDF online