S1716
Physics - Dose prediction/calculation, optimisation and applications for particle therapy planning
ESTRO 2026
beam angle selection as a 72-class multi-label classification problem (5° steps). Angular periodicity was managed using a circular smooth label technique. Validation was performed with 5-fold cross- validation.The DL-based angle prediction outputs a 72- dimensional probability vector corresponding to 5° steps for the test cases. Post-Processing was implemented to refine the initial DL probability output. A cost-function was used penalizing beam angles traversing critical Organs-at-Risk (OARs), such as the spinal canal or heart. The results from the Post- Processing were considered as the final DL beam angles.Final DL beam angles were geometrically compared to the human selected ground truth beam angles. Then, treatment plans were generated for both sets of angles using an automated knowledge-based planning procedure. Resulting plans were normalized to the same Clinical Target Volume (CTV) coverage (V100%=95% for 6000cGy in 30 fractions). A dosimetric comparison was performed for Esophagus, Heart, Lungs-CTV, and Spinal Canal using a paired t-test (p<0.05). Results: DL angles were either very similar to ground truth or represented different but clinically valid options (Figures 1 and 2). Dosimetric analysis showed no significant (p<0.05) dosimetric differences between plans generated from ground truth and DL beam angles. OAR metrics were comparable: Esophagus Dmax=3940±2695cGy (mean±standard deviation) for ground truth plans versus 4080±2699cGy for DL plans. Heart Dmean 450±684cGy vs. 390±600cGy. Lungs-CTV Dmean 785±941cGy vs. 832±978±cGy. Spinal Canal Dmax 2053±2012cGy vs. 2490±1955cGy.
Conclusion: The proposed DL solution predicts valid beam angles for lung IMPT, resulting in clinically acceptable plans. Employing this algorithm can increase clinical efficiency. Future work will focus on further reducing OAR dose. Keywords: lung cancer, IMPT beam angles, deep- learning Mini-Oral 1344 Optimizing range shifter use to reduce intrauterine doses in pencil beam scanning proton therapy during pregnancy Jana Hohmann 1 , Anneleen Goedgebeur 2 , Marijke De Saint-Hubert 3 , Dries Colson 2 , Frédéric Amant 4,5 , Maarten Lambrecht 1,2 , Tom Depuydt 1,2 1 Department of Oncology, KU Leuven, Leuven, Belgium. 2 Department of Radiation Oncology, UZ Leuven, Leuven, Belgium. 3 Radiation Protection Dosimetry and Calibration Expert Group, Belgian Nuclear Research Center (SCK CEN), Mol, Belgium. 4 Department of Obstetrics and Gynaecology, Division of Gynaecological Oncology, KU Leuven, Leuven, Belgium. 5 Department of Oncology, Unit of Gynaecological Oncology, UZ Leuven, Leuven, Belgium Purpose/Objective: In radiotherapy during pregnancy, out-of-field (OOF) doses should be as low as reasonably achievable, with fetal exposure below 100 mSv to prevent deterministic effects and ideally near public dose limits to minimize stochastic risk. Proton therapy with pencil beam scanning (PBS) supports these objectives by delivering lower OOF dose than photon radiotherapy [1].
Made with FlippingBook - Share PDF online