S2477
Physics - Radiomics, functional and biological imaging, and outcome prediction
ESTRO 2026
was defined using the same ADC band (800–1100 × 10 ⁻ ⁶ mm ² /s) from [1], and the optimal cut-off value for classification (5.8 cm ³ ) was also the same reported in [1]. We applied logistic regression to model the association between HRS and progression-free survival (PFS) up to 5 years. To test robustness of HRS against GTV segmentation variability, a sensitivity analysis was performed by introducing random perturbations to the GTVs. Following the augmentation method [2], each axial slice was independently displaced within predefined bounds, and out-of-plane variability was simulated by randomly shifting the first and/or last slice by one slice.We also investigated the discriminative ability of the morphological tumor volume (GTVmorph) and of the ratio HRS/GTVmorph. Results: Progression of disease was registered in 7/34 patients. The HRS biomarker reproduced the expected prognostic behavior (Figure_1A), with an AUC=0.79, accuracy=0.82 (Fisher’s exact p-value<0.01). The GTVmorph alone also showed similar discriminative power (AUC=0.78), consistent with the findings in [1]. A combined model including the GTVmorph and HRS/GTVmorph dichotomised at 0.25 yielded AUC=0.85 and increase in precision-recall AUC from 0.72 to 0.75, confirming the independent role of HRS in addition to GTVmorph.Kaplan–Meier analysis (Figure_1B) confirmed statistically significant stratification between the two groups (below/above 5.8 cm ³ cutoff for HRS, p=0.005). Perturbation analysis of GTV contours demonstrated stable biomarker performance, with unchanged AUC values, confirming the model's robustness to segmentation variability.
those with higher MDC. The majority of radiomic features are repeatable throughout treatment. These findings highlight the importance of considering the temporal behaviour of radiomic features across treatment, rather than relying on features derived from baseline imaging. References [1] Shakur, A., Lee, J.Y.J. and Freeman, S. (2023). An Update on the Role of MRI in Treatment Stratification of Patients with Cervical Cancer. Cancers, 15(20), p.5105. [online]. Available from: http://dx.doi.org/10.3390/cancers15205105. [2] O'Connor, J., Aboagye, E., Adams, J. et al. Imaging biomarker roadmap for cancer studies. Nat Rev Clin Oncol 14 , 169–186 (2017). https://doi.org/10.1038/nrclinonc.2016.162 Mini-Oral 3891 Independent external validation of an ADC-based prognostic biomarker for outcomes after radiotherapy in head-and-neck cancer Enrico Longhi 1 , Eliana Gioscio 1 , Nicola Alessandro Iacovelli 2 , Marzia Franceschini 2 , Marco Dassie 3 , Anna Cavallo 4 , Giuseppina Calareso 3 , Ester Orlandi 5,6 , Tiziana Rancati 1 1 Data Science Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy. 2 Radiotherapy Department, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy. 3 Department of Radiology, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy. 4 Medical Physics, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy. 5 Radiation Oncology Clinical Department, National Center for Oncological Hadron Therapy (CNAO), Milan, Italy. 6 Department of Clinical, Surgical, Diagnostic, and Pediatric Sciences, University of Pavia, Pavia, Italy Purpose/Objective: Winter_2025 [1] identified MRI-derived high-risk subvolumes (HRS) based on apparent diffusion coefficient (ADC) from diffusion-weighted imaging (DWI) as predictive of radiotherapy (RT) failure in head- and-neck cancer (HNC) patients. In the present study, we aimed to validate this prognostic biomarker for RT outcome in an external independent HNC cohort. Material/Methods: We retrospectively computed HRS for 34 HNC patients (enrolled in a prospective study for the development of predictive models of radiotherapy outcomes) who underwent pre-treatment DWI and received radical RT (69.96Gy over 33 fractions). Table_1 reports comparison of clinical characteristics of our cohort vs Winter’s. Gross Tumor Volumes (GTVs) were delineated on T2-weighted images and co-registered to the corresponding ADC maps. For each patient, HRS
Conclusion: Despite some differences in cohorts, this validation study confirms the prognostic value of the ADC- defined HRS biomarker for RT outcome in HNC (with
Made with FlippingBook - Share PDF online