S675
Clinical – Head & neck
ESTRO 2026
Results: Non-rigid registration to the common coordinate system of the mandible has a mean Dice score of 87.7 ± 1.5 and an Average Surface Distance of 0.40 ± 0.05 mm. In Group 2, the mean dose for the non-toxic subregion was 29.87 Gy, while for the toxic subregion it was 53.64 Gy. In Group 4, the mean dose for the non-toxic subregion was 35.03 Gy, and for the toxic subregion it was 53.63 Gy. Figure 1 shows anatomical subregions associated with ORN toxicity and corresponding DVHs.Figure 1 Anatomical subregions associated with ORN toxicity and their corresponding DVH.
Digital Poster 5187 Study of dental osteoradionecrosis after head and neck radiotherapy, a voxel-based analysis for identifying local dose-effect relationships Lisset Zea 1 , Grégory Delpon 2 , Lucía Cubero 3,1 , Anne- Gaëlle Chaux 4 , Joel Castelli Castelli 1 , Margot Métivier 2 , Franck Jegoux 5 , Olivier Malard 4 , Renaud Crevoisier 1 , Javier Pascau 1 , Juliette Thariat 6 , Mélanie Doré 2 , Oscar Acosta 1 1 INSERM - LTSI - UMR 1099, Univ Rennes - CLCC Eugène Marquis, Rennes, France. 2 ICO, Institut de Cancérologie de l’Ouest, Nantes, France. 3 Departamento de Bioingeniería, Universidad Carlos III de Madrid, Madrid, Spain. 4 Nantes Université, CHU Nantes, Nantes, France. 5 Department of Radiation Oncology,, Centre François Baclesse, Rennes, France. 6 Department of Radiation Oncology,, Centre François Baclesse, Caen, France Purpose/Objective: Osteoradionecrosis of the mandible is a radioinduced complication for patients with head and neck cancer which can lead to severe functional sequelae. Developing strategies to predict osteoradionecrosis is crucial in order to improve patients' quality of life. While dosimetric risk factors in the the Intensity-Modulated Radiation Therapy(IMRT) era remain controversial, current studies rely on region-based analyses Most current approaches have focused on region-based methods to investigate the relationship between radiation dose and osteoradionecrosis [1]. In this study, we propose a voxel-based analysis (VBA)[2] to unveil local dose- effect relationships, aiming to identify mandibular subregions where the dose predicts toxicity. Material/Methods: A cohort of 99 patients treated with IMRT for carcinoma of the oral cavity (OC) or oropharynx (OR), of which 50 patients had ORN and 49 were in the control group. Jaw segmentation was performed using deep learning-based segmentation [3]. Each patient was aligned to a common coordinate system using a non-rigid multi-step registration process, guided by segmentation Euclidean distance maps. Deformation fields were used to propagate 3D dose distributions into the reference space.Four groups were created: (1) all patients with ORN and controls, (2) patients with central tumors and ORN, (3) patients with ORN in regions 35-38 or 45-48 with ipsilateral tumor laterality, and (4) patients with ORN in the same regions but contralateral. VBA was performed on non- rigidly aligned dose distributions. Voxel-wise Mann- Whitney U-test (p < 0.01) identified significant voxels, and a non-parametric permutation test (1000 permutations) corrected for multiple comparisons. Significant subregions were found from both tests.
Figure 1. Anatomical subregions predictive and DVH. Conclusion: VBA enabled the identification of anatomical subregions where the dose distribution correlates with ORN. Mean doses in these subregions were significantly higher and could be used in predictive models. These findings pave the way for personalized irradiation planning. References: [1] Renouf M, Auger et al. . Prognostic Factors of Mandibular Osteoradionecrosis Including Accurate Colocalization of Avulsions and Dosimetric Dental Mapping Software, a Case-control Study. Int J Radiat Oncol Biol Phys. 2024 Apr 27:S0360-3016(24)00518- 2.[2] Sosa-Marrero, C., Acosta, O., et al. Voxel-wise analysis: A powerful tool to predict radio-induced toxicity and potentially perform personalised planning in radiotherapy. Cancer/Radiothérapie, (2023). 27(6-7), 638-642. [3] Cubero, L., Hémon, et al. Deep learning- based segmentation of head and neck organs at risk on CBCT images with dosimetric assessment for radiotherapy. Physics in Medicine & Biology, (2025). 70(7), 075014. Keywords: Voxel based analysis, Osteoradionecrosis
Made with FlippingBook - Share PDF online