S1974
Physics - Dose prediction/calculation, optimisation and applications for photon and electron planning
ESTRO 2026
Pesznyák 1,2 , Gábor Stelczer 1,2 , Péter Ágoston 1,4 , Zoltán Takácsi-Nagy 1,4 , Tibor Major 1,4 1 Centre of Radiotherapy, National Institute of Oncology, Budapest, Hungary. 2 Institute of Nuclear Technique, Budapest University of Technology and Economics, Budapest, Hungary. 3 Doctoral College, Semmelweis University, Budapest, Hungary. 4 Department of Radiotherapy, Semmelweis University, Budapest, Hungary
Minor deviations were observed for specific rectal dose constraints and whole body low dose exposure. Further investigation of planning priorities sequence is warranted in order to achieve full clinical compliance. Keywords: ETHOS, prostate, dosimetric Poster Discussion 4032 A fast Monte Carlo engine for microbeam radiotherapy dose calculation Liam Day 1 , Ondar Skrla 2 , Johanna Winter 1 , Stefan Bartzsch 1 , Stephanie E Combs 1,3 , Kim M Kraus 1 1 TUM School of Medicine and Health, Department of Radiation Oncology, TUM University Hospital, Technical University of Munich, Munich, Germany. 2 TUM School of Natural Sciences, Technical University of Munich, Munich, Germany. 3 Partner Site Munich,
Purpose/Objective: To investigate the feasibility of stereotactic
radiotherapy planning for prostate cancer patients using the Varian Ethos system, regarding target coverage and dosimetric parameters of organs at risk. Material/Methods: For our dosimetric study, we selected ten prostate cancer patients previously treated at our institution using CyberKnife (CK). The original treatment plans were created with a simultaneous integrated boost technique, where the volume including the prostate and seminal vesicles (PVS) received 5 × 6 Gy, while the prostate (PROS) received 5 × 7.5 Gy. The CTV to PTV margin for CK was 3 mm. The PTVs were created with 5 mm safety margin on Ethos, as continuous tracking of the target is not possible. The Ethos system includes an AI-assisted intelligent optimization engine to perform optimization based on clinical goals set for the treatment. Preset beam geometries can be selected in the system; for comparison, we chose 9- and 12-field IMRT, and VMAT plans with 2 or 3 full arcs. We analyzed the performance of the Ethos planning system and compared the plans with the CK plans. Dosimetric data (i.e. V95, V100) were collected for the target volumes (CTV, PTV) and for the main organs at risk (bladder wall, rectum, sigmoid colon, whole body). Results: Target coverage for PVS was adequate in all Ethos- generated plans. The V100 values were on average 0.3% lower than the CK values, and the V95 values 0.7% lower, (99.52% and 99.95%, respectively). In IMRT plans, the PROS V95 values were 1% lower than in CK plans, while the VMAT plans achieved similar or 0.5% better results, on average. CTV coverage was adequate in all cases. Unlike the CK plans, the Ethos plans could not fully meet the D0.04ccm dose constraint for the rectum ( ≤ 38 Gy), exceeding it by 0.1–0.2 Gy. The bladder wall D0.04ccm value ( ≤ 40 Gy) was within limits for IMRT plans, while VMAT plans slightly exceeded the constraint. For the sigmoid colon, both evaluated parameters were within limits for all plans. The whole- body V15Gy values were on average 76% higher compared to CK, and were higher for IMRT than for VMAT among the Ethos plans. Conclusion: AI-based Ethos planning can produce clinically acceptable stereotactic prostate treatment plans.
Deutsches Konsortium für Translationale Krebsforschung (DKTK), Munich, Germany
Purpose/Objective Microbeam radiotherapy (MRT) is a promising technique for cancer treatment but has remained in a preclinical state. Dose calculation for MRT is extraordinarily challenging and relies on time intensive MC simulations. A fast MC engine for MRT dose calculation is required to meet clinical standards. Extensions to the algorithm are also required to compute the relevant dosimetric information unique to MRT. Material/Methods A custom MC engine was written for the hybrid dose calculation algorithm 1 . The Geant4 v11.3.0 Livermore physics models and cross-section data 2,3 were adapted for the new engine. The MC engine was ported to the GPU. In addition to calculating the dose-to-medium (D m ), the hybrid algorithm was extended to compute the dose- to-water (D w ) and the dose-to-marrow/tissue (D t ) 4 . The hybrid algorithm was paired with the Eclipse TPS v16.1 (Varian, Inc.) using the Eclipse Algorithm API. Results The updated hybrid-MC (CPU) engine running on an Intel Xeon Silver 4110 processor improved calculation times by a factor of ~16x compared to the hybrid- Geant4 implementation. The hardware accelerated hybrid-MC (GPU) running on an NVIDIA RTX A5000 improved calculation times by a factor of ~26x compared to hybrid-MC (CPU); a ~420x improvement compared to hybrid-Geant4. The hybrid-MC (GPU) algorithm takes approximately 5 minutes to calculate per field for the test case shown in figure 1 (2e9 histories per field), while hybrid- Geant4 calculations take approximately 36 hours per field.
Made with FlippingBook - Share PDF online