ESTRO 2025 - Abstract Book

S2925

Physics - Dose prediction, optimisation and applications of photon and electron planning

ESTRO 2025

mm MLC demonstrated superior dose gradients, better conformity, and lower doses to OARs compared to plans with 5 mm MLC. The Varian Edge system with 2.5 mm MLC is recommended for complex spine SABR cases.

Keywords: Spine SABR - Eclipse – Elements

References: Sahgal, A., et al. (2019). Stereotactic body radiotherapy for spinal metastases: A review. Journal of Clinical Oncology, 37(30), 2798–2806. Benedict, S. H., Yenice, K. M., Followill, D., et al. (2010). Stereotactic body radiation therapy: The report of AAPM Task Group 101. Medical Physics, 37(8), 4078–4101. Ryu, S., et al. (2017). International consensus on the diagnosis and treatment of spine metastases. Lancet Oncology, 18(12), e697-e706. Ong, C. L., Verbakel, W. F., Cuijpers, J. P., et al. (2011). Stereotactic radiotherapy for spine metastases using volumetric intensity-modulated arc therapy: Dosimetric and clinical evaluation. International Journal of Radiation Oncology Biology Physics, 83(5), e805-e812.

4193

Mini-Oral Radiotherapy personalization through risk-adjusted multi-criteria optimization Katrin Teichert 1 , Mara Schubert 1 , Zhongxing Liao 2 , Thomas Bortfeld 3 , Ali Ajdari 3

1 Optimization, Fraunhofer ITWM, Kaiserslautern, Germany. 2 Department of Radiation Oncology, University of Texas‘ MD Anderson Cancer Center, Houston, USA. 3 Department of Radiation Oncology, Massachusetts General Hospital & Harvard Medical School, Boston, USA Purpose/Objective: With growing clinical data records and advances in machine learning methods, predicting the risk of adverse side effects from patient individual characteristics becomes more practicable (Isaksson, et al. 2020). If those learned models also incorporate dose related risk factors, they hold the potential to customize a plan to a patient’s unique risk profile (Maragno, et al. 2024). We propose a new method for including learned risk prediction models into multi-criteria plan optimization (MCO) (Craft, et al. 2012). The goal of our proposed approach is to derive a set of patient-specific risk-optimized counterparts to the MCO-generated clinical plans, with minimal deviation in the clinical objectives. Material/Methods: We implemented a new method to calculate a set of treatment plan alternatives. Therein, standard objectives derived from clinical protocols are considered of prime importance and optimized to their full potential within the specified constraints. The risk models are considered of secondary importance, and should only be improved as long as each trade-off to a primary objective does not exceed a specified threshold. To achieve this in a single MCO calculation (rather than a stepwise approach), a specifically modified order relation is used within the MCO sandwiching algorithm (Bokrantz 2013). This allows us to retain all the algorithm’s benefits in terms of calculation efficiency and approximation accuracy. We tested our methodology for several models predicting both RT-induced cardiovascular disease (RICD) and radiation pneumonitis (RP). The models were trained on a non-small cell lung cancer (NSCLC) RT dataset (n=179) treated with photon (IMRT) and protons, and then integrated within MCO-based treatment planning. The difference between the conventional (risk-agnostic) and the proposed methodology (risk guided MCO) were quantified in terms of dosimetric criteria as well as the predicted RICD and/or RP. Results: We applied our method to generate personalized, risk-guided plan alternatives for 7 NSCLC patients based on their predicted risk and within clinical criteria. For a risk model associating increased total lung V20 with higher risk for