S1932
Physics - Dose prediction/calculation, optimisation and applications for photon and electron planning
ESTRO 2026
Results: Automated GTV segmentation was unsuccessful (leading to a DSC = 0) in two patients due to the small tumor size and the post-processing step that currently is tailored to select only the largest tumor when multiple tumors are present. These cases were, therefore, excluded from the comparisons. The average ± standard deviation DSC between the original and the automated GTVs was 0.83 ± 0.10. Dose differences between the original and replanned treatments were minimal: The D95% of the PTV was comparable between plans (60Gy for both). Though fulfilling all clinical constraints, the mean doses to the heart (8.9 vs. 7.6Gy, p<0.05), esophagus (7.1 vs. 6.8Gy), and lungs (12 vs. 11Gy) were increased nominally for the replanned treatments. Table: Dosimetric comparison between the original and replanned ECHO plans. Asterisk denotes statistically significant differences (Wilcoxon signed- rank test). Conclusion: This proof-of-concept study demonstrates that replanning based on AI auto-segmented GTVs is feasible, and results in high geometric agreement with manual GTVs and associated negligible differences in tumor and normal tissue dose. As a next step, we plan to implement quality assurance for auto-segmented contours, and additional studies are underway to evaluate its impact on plan quality. References: [1]J. Jiang, N. Tyagi, K. Tringale, C. Crane, and H. Veeraraghavan, “Self-supervised 3D Anatomy Segmentation Using Self-distilled Masked Image Transformer (SMIT),” in Medical Image Computing and Computer Assisted Intervention – MICCAI 2022: 25th International Conference, Singapore, September 18– 22, 2022, Proceedings, Part IV, Berlin, Heidelberg: Springer-Verlag, Sept. 2022, pp. 556–566. doi: 10.1007/978-3-031-16440-8_53. [2]M. Zarepisheh et al., “Automated intensity modulated treatment planning: The expedited constrained hierarchical optimization (ECHO) system,” Med Phys, vol. 46, no. 7, pp. 2944–2954, July 2019, doi: 10.1002/mp.13572. Keywords: Autoplanning, Autocontouring, Esophageal Cancer
Digital Poster 3337 Convert-then-map vs map-then-convert: impact on mapped dose assessment in reirradiation Chelmis Muthoni Thiong'o 1 , Marcel van Herk 1,2 , Matthew Lowe 1,3 , David Thomson 1,4 , Clara Chan 4 , Ane Appelt 5 , Eliana Vasquez Osorio 1,2 1 Division of Cancer Sciences, The University of Manchester, Manchester, United Kingdom. 2 Radiotherapy-Related-Research Group, The Christie Hospitals NHS Foundation Trust, Manchester, United Kingdom. 3 Department of Medical Physics and Engineering, The Christie Hospitals NHS Foundation Trust, Manchester, United Kingdom. 4 Department of Clinical Oncology, The Christie Hospitals NHS Foundation Trust, Manchester, United Kingdom. 5 Department of Oncology, Rigshospitalet, Technical University Hospital of Greater Copenhagen, Copenhagen, Denmark
Purpose/Objective: In reirradiation, cumulative dose assessment
requires i) converting physical dose to equieffective dose, and ii) mapping previous dose distributions onto current anatomy. As dose conversion is non-linear and dose mapping uses linear interpolation, the sequence of these steps may influence the mapped dose distributions with no consensus on step order ¹ . We assessed the impact of the conversion-mapping sequence order in two reirradiation patient cohorts, with different fractionations. Material/Methods: We used data from 54 patients who underwent head and neck (HN) or lung reirradiation. Intensity-based deformable registration was performed using default settings (RayStation, v11B-R, RaySearch Lab, Stockholm, Sweden), with the reirradiation scan as fixed image.Two dose mapping approaches were evaluated:Convert-then-Map (CM), where previous dose distribution was converted to Equieffective Dose in 2 Gy fractions (EQD2), then mapped onto the reirradiation scan.Map-then-Convert (MC), where the previous dose distribution was first mapped, then converted to EQD2.The same α / β value was applied to the entire dose distribution to avoid artefacts at organ at risk boundaries, Figure 1a. We tested α / β = 1, 2, and 3 Gy.To assess the impact of each sequence, voxel- wise CM-MC differences were computed. We recorded the extreme negative and positive voxel-wise differences across complete distributions for each patient. Then, aggregated analyses were performed by characterising CM–MC differences using median and interquartile summaries within 5% intervals of the prescribed dose (Figure 1b). Results: Aggregated analysis revealed a clear pattern, where the largest differences were observed at ~80%
Made with FlippingBook - Share PDF online