ESTRO 2025 - Abstract Book

S3985

Radiobiology - Tumour radiobiology

ESTRO 2025

1 Department of Radiation Oncology, Sun Yat-sen University Cancer Center, Guangzhou, China. 2 Collaborative Innovation Center for Cancer Medicine, Guangdong Key Laboratory of Nasopharyngeal Carcinoma Diagnosis and Therapy, State Key Laboratory of Oncology in South China, Guangzhou, China. 3 Research cooperation department, Shenzhen United Imaging Research Institute of Innovative Medical Equipment, Shenzhen, China. 4 Zhongshan School of Medicine, Sun Yat-sen University, Guangzhou, China. 5 Imaging Physics Algorithm Department, Shanghai United Imaging Healthcare Co., Ltd., Shanghai, China Purpose/Objective: PET imaging is pivotal in visualizing the high metabolic regions (HMRs) of nasopharyngeal carcinoma (NPC). However, its associated radiation exposure and high costs limit its application in real-time monitoring during radiotherapy. Therefore, our study aims to automatically delineate HMRs on Computed Tomography (CT) scans without relying on PET inputs, providing a noninvasive and cost-effective tool for the biology-guided adaptive radiotherapy (ART). Material/Methods: We retrospectively collected 18 F-FDG PET/CT images from 399 patients (with 40 cases in the test set) before radiotherapy and from 3 patients during radiotherapy as an additional test set at a single medical center in 2021. An automated pipeline was designed to extract HMRs from PET images as labels (Figure 1). HMRs were defined as areas within the primary gross tumour volume (GTVp) with standardized uptake values (SUV) greater than 2.5, excluding any overlap with bone, air, or brain tissue. After cropping around the GTVp center, we developed a seven layer 3D MedNext model with dual-channel input (Figure 1) and benchmarked its performance against the baseline nnUNetV2. Our innovation lies in eliminating the dependence on PET inputs and combining the Dice coefficient with focal loss as the loss function for precise delineation of small-volume samples. The average Dice similarity coefficient (DSC) and average symmetric surface distance (ASSD) were used as comprehensive performance metrics both before and during radiotherapy. Additionally, we examined predicted volumetric alterations in periodic fan beam computed tomography (FBCT) scans (before the 23rd fraction) across three patients undergoing radiotherapy to test the robustness.