S2425
Physics - Radiomics, functional and biological imaging, and outcome prediction
ESTRO 2026
Digital Poster Highlight 1173 Rethinking radionecrosis risk with a tissue-based dose–volume perspective beyond V12Gy Szabolcs David 1,2 , Laura Koldenhof 1 , Daniel C.A. Versnel 1 , Marielle Philippens 1 , Joost J. C. Verhoeff 1,2 , Enrica Seravalli 1 1 Department of Radiation Oncology, University Medical Center Utrecht, Utrecht, Netherlands. 2 Department of Radiation Oncology, Amsterdam University Medical Center, Amsterdam, Netherlands Purpose/Objective: Radiation necrosis (RN) after stereotactic radiotherapy (SRT) for brain metastases (BM) is a clinically significant late toxicity that compromises neurological function and quality-of-life[1]. MR - Linac SRT for BM may streamline care and logistics, but its coplanar delivery can modestly increase low - dose bath. To minimize RN risk, planning constraints often emphasize dose–volume metrics, most notably V12Gy[2], despite heterogeneous findings and reproducibility issues across studies. Moreover, these recommendations[3], [4] do not consider the underlying tissue and cerebrospinal fluid (CSF) is included in volumetrics, which has no relevance to parenchymal injury, diluting dose–effect signals. In this work, we study RN risk per tissue class (gray matter (GM), white matter (WM), and CSF) using irradiated isodose volumes across a wide dose range (V1–V25Gy) to determine dose–volume thresholds most strongly A retrospective cohort of 244 patients with single- lesion BM treated with LINAC-based SRT (2012–2023) was selected. The median age was 65, 51% were female, 87 patients developed RN, and the median GTV was 7.7 cm3; 77% received single-fraction non- coplanar VMAT Linac-based SRT, and multi-fraction doses were converted to single-fraction equivalent dose using α / β =2 Gy. Other treatments included surgery (38%), immunotherapy (37%), and chemotherapy (37%). Planning (pre-SRT) T1w MRIs were segmented into WM, GM, and CSF using CAT12[5]; due to frequent misclassification of WM with associated with RN. Material/Methods: edema into GM, a combined ‘tissue-only’ GM+WM class was defined. V1–V25Gy was computed from the dose distribution. Logistic regression (1-univariable; 2- age/sex-adjusted; 3-treatment-adjusted (surgery, chemotherapy, immunotherapy); 4-fully adjusted) related isodose volumes to RN presence (1-vs-0). Figure 1 illustrates the segmentation and tissue- specific dose-volume setup.
Results: In univariate tissue-only modeling, only V18–V21Gy were significant for predicting RN risk, odds ratio (OR): 1.082/1.084/1.13/1.16 for V18/V19/V20/V21, respectively. The fully adjusted tissue-only model identified significantly V14–V21Gy, OR: 1.028/1.032/1.06/1.08/1.107/1.11/1.17/1.19 for V14/V15/V16/V17/V18/V19/V20/V21, respectively. Figure 2 shows two examples of the regression and the resulting OR per dose-volume threshold plot for univariate and fully adjusted (demography+treatment) models.
Made with FlippingBook - Share PDF online