ESTRO 2025 - Abstract Book

S3035

Physics - Image acquisition and processing

ESTRO 2025

3236

Digital Poster Accuracy of deformable dose mapping in thoracic reirradiation Nicholas Hardcastle 1,2 , Enar Recalde Vizcay 3 , Susan Harden 4 , Adam Yeo 1

1 Physical Sciences, Peter MacCallum Cancer Centre, Melbourne, Australia. 2 Sir Peter MacCallum Department of Oncology, University of Melbourne, Melbourne, Australia. 3 Radiation Oncology, Vall d'Hebron University Hospital, Barcelona, Spain. 4 Radiation Oncology, Peter MacCallum Cancer Centre, Melbourne, Australia Purpose/Objective: Safe reirradiation requires adherence to critical organ cumulative dose constraints [1]. There is wide variability in methods to estimate cumulative dose to critical organs. Deformable dose mapping (DDM) is more efficient and consistent compared with other approaches [2] though the question of accuracy of DDM still remains. We estimate the accuracy of DDM for thoracic reirradiation through systematic analysis of dose to corresponding anatomical landmarks. Material/Methods: Ten patients with lung cancer treated with repeat courses of thoracic radiation therapy were included. Patients were Type-1 reirradiation [3], receiving cumulative BED10 prescription doses between 103-144 Gy. Ethics approval was obtained. A radiation oncologist identified 15-25 corresponding anatomical landmarks in the irradiated volume on each pair of treatment planning CTs. Rigid (RIR) and three deformable image registration (DIR) algorithms from the VelocityAI software (v4.1, Varian Medical Systems, Palo Alto, USA) were used to perform DDM from the Course 1 to the Course 2 planning CT. DIR algorithms were deformable multi-pass (DMP), extended- DMP (EDMP) and structure-guided (SG), for which all landmarks were used as guidance structures. In this SG algorithm, structures do not act as exact boundary conditions, but cost function weighting is increased at the region of the included structure. Target registration error (TRE) was computed using deformed landmarks. Absolute difference in dose to corresponding landmark in Course 1 image, and from the deformed dose at that landmark in the Course 2 image was computed as dose mapping error (DME). Mann-Whitney test was used to compare distributions of TRE and DME over the distribution of all points from all patients. Results: Median (IQR) TRE was reduced from 6.1 (3.9-9.7) mm with RIR to 2.6 (1.9-4.2) mm with DMP, 2.6 (1.7-4.0) mm with EDMP and 2.6 (2.0-4.6) mm with SG (all p<0.005) (Figure 1a). The Median (IQR) absolute value of the DME was reduced from 0.4 (0.1-2.0) Gy with RIR to 0.3 (0.1-0.7) Gy with DMP, 0.2 (0.0-0.7) Gy with EDMP and 0.1 (0.0-0.5) Gy with SG (all p<0.005) (Figure 1b). Except for three points in very high dose gradient at field edge in two patients, DDM for all DIR algorithms was <10 Gy.