S1931
Physics - Dose prediction/calculation, optimisation and applications for photon and electron planning
ESTRO 2026
therapy, Radiation Oncology, doi:10.1186/s13014-016- 0747-y. 4. Kerkmeijer, L.G.W. et al. (2018) Magnetic resonance imaging only workflow for radiotherapy simulation and planning in prostate cancer , doi:10.1016/j.clon.2018.08.009. 5. Nyholm, T. et al. (2018) MR and CT data with multiobserver delineations of organs in the pelvic area- Part of the Gold Atlas project , doi:10.1002/mp.12748. A feasibility study of an end-to-end automated planning workflow for lung cancer RT: AI-based auto-contouring and optimization-driven auto- planning Chloe Min Seo Choi 1 , Jonas Willmann 2 , Nikhil M Mankuzhy 3 , Gourav Jhanwar 1 , Sharif Elguindi 1 , Masoud Zarepisheh 1 , Jue Jiang 1 , Maria Thor 1 , Harini Veeraraghavan 1 1 Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, USA. 2 Department of Radiation Oncology, University Hospital Zurich, Zurich, Switzerland. 3 Department of Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York, USA Purpose/Objective: The purpose of this feasibility study was to evaluate an end-to-end automated contouring and radiotherapy (RT) planning framework and to determine whether treatment plans derived from auto-segmented tumors achieve comparable plan quality based on manually Digital Poster 3335
reference CTs (mean MAE 44.2 HU, PSNR = 28.1dB, and SSIM 0.91). Dose recalculations showed excellent agreement (gamma pass rate 98.8% (1%/1mm) and 99.7% (2%/2mm), MADD = 0.017 Gy, and 96.6% of voxels within ±0.1 Gy) . All DVH endpoint differences were clinically negligible.
contoured tumors. Material/Methods: Ten patients with lung cancer without nodal
involvement who previously underwent intensity- modulated RT were included. For each patient, the planning CT and the corresponding structure set were collected. The gross tumor volume (GTV) was auto- segmented on the planning CT using a previously published AI-based contouring model [1]. To evaluate the impact of using auto-contoured targets, treatment replanning was performed with an in-house developed hierarchical optimization-based automated algorithm (ECHO), using the same planning objectives and optimization parameters as the clinically delivered ECHO plans [2]. The manually contoured and auto- contoured GTVs were compared geometrically using the Dice similarity coefficient (DSC). Dose comparisons were performed between the original treatments and the re-planned treatments for the planning target volume (PTV; created by adding an isotropic margin of 12mm to the GTV) and the heart, esophagus, and lungs using the Wilcoxon signed-rank test.
Conclusion The evaluated MR-only AI-workflow demonstrated geometric, image, and dose accuracy comparable to expert manual and CT-based references. Small systematic deviations, particularly for the penile bulb, likely reflect definitional differences rather than geometrical errors. Overall, the results confirm that AI- driven segmentation and MR-sCT synthesis provide clinically robust performance for prostate and pelvic radiotherapy planning, supporting their integration into routine MR-only treatment workflows. References 1. Owrangi, A.M. et al. (2018) MRI-only treatment planning: Benefits and challenges, doi:10.1088/1361- 6560/aaaca4. 2. Jonsson, J. et al. (2019) The rationale for MR-only treatment planning for external radiotherapy , doi:10.1016/j.ctro.2019.03.005. 3. Edmund, J.M. and Nyholm, T. (2017) A review of substitute CT generation for MRI-only radiation
Made with FlippingBook - Share PDF online