S2781
RTT - RTT contouring, target definition, and treatment planning
ESTRO 2026
Results: Median (IQR) for TRE, DME and MSD all decreased with deformable registration. For both cohorts, SG_LM achieved the lowest median TRE and DME at the landmarks, while DMP produced the lowest median MSD. These results are summarised in Table 1.
no further adjustments in the treatment planning approach were needed. Keywords: SBRT, Ethos, treatment planning
Digital Poster Highlight 2521
Toward Improving Treatment Planning Accuracy for Reirradiation: Assessment of Deformable Dose Mapping for Thoracic and Head and Neck Patients Catherine Laferlita 1,2 , Brett Clark 1,3 , Katrina Woodford 1,2 , Kenton Thompson 1,2 , Tsien Fua 1,2 , Arnulf Mayer 1 , Julianne O'Shea 1 , Adam Yeo 1,2 , Nicholas Hardcastle 1,2 1 Department of Radiation Oncology, Peter MacCallum Cancer Centre, Melbourne, Australia. 2 Sir Peter MacCallum Department of Oncology, The University of Melbourne, Melbourne, Australia. 3 Image X Institute, The University of Sydney, Sydney, Australia Purpose/Objective: Advances in anticancer therapies have improved patient survival, leading to an increasing number of patients requiring additional radiation therapy to the same or adjacent anatomical sites, a practice known as reirradiation (reRT) ¹ . Accurate dose summation in reRT is critical: underestimation risks overdosing healthy tissue, whereas overestimation may compromise treatment efficacy. This retrospective study aims to evaluate the accuracy of deformable dose mapping (DDM) in thoracic and head & neck reRT patients. Material/Methods: Forty reRT patients (20 thoracic, 20 head & neck) were analysed. For each case, 15-25 anatomical landmarks were identified on the planning computed tomography (pCT) for each course. Contour+ (MVision, Helsinki, Fi) was used for contour generation across all datasets. Rigid image registration (RIR) and three deformable image registrations (DIR) – deformable multi-pass (DMP), extended deformable multi-pass (EDMP), and structure-guided using identified landmarks (SG_LM) – were performed in Velocity AI (v4.2, Varian Medical Systems, Palo Alto, USA) to enable a subset of critical structures and dose to be mapped from initial course (C1) to reRT course (C2). Registration accuracy was evaluated using Target Registration Error (TRE) at deformed landmarks. Dose Mapping Error (DME) quantified differences between the doses at each landmark. Contour alignment was assessed using Mean Surface Distance (MSD). Consistency in near-maximum dose for organs of interest was evaluated as the difference between D0.03cc computed in C1 and that in C2 after dose mapping. Median TRE, DME and MSD were compared between deformable and rigid methods using a Wilcoxon signed-rank test (p < 0.05).
Figure 1 illustrates the differences in dose metrics between C1 ground truth and C1 dose mapped to the same structure in C2. For most structures (8 out of 12) DMP or EDMP showed the lowest median dose error, whilst for the vast majority (10 out of 12) RIR or SG_LM showed the highest.
Conclusion: More precise cumulative dose analysis with DDM has potential to improve treatment planning accuracy for reRT patients. Incorporating anatomical landmarks in regions of concern can further enhance accuracy. For the algorithms evaluated, DMP and EDMP were favourable for borders of organs of interest. Individual case review remains essential to identify potential dose mapping errors. References:
Made with FlippingBook - Share PDF online