Integrating Intratumoral and Peritumoral 18F-FDG PET/CT Radiomics for Overall Survival Prediction in Head and Neck Cancer: A Multicenter Study

Objective: This study aimed to develop a nomogram integrating intratumoral and peritumoral radiomics from pretreatment 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography/computed tomography (PET/CT) to predict overall survival (OS) in head and neck cancer (HNC). Methods: This multicenter retrospective study included 758 pathologically confirmed HNC patients. The primary tumor was segmented and expanded by 1–5 mm to generate multi-scale peritumoral regions. Radiomics features were extracted from intratumoral and peritumoral volumes on PET and CT, followed by feature selection. Multiple survival models were constructed and compared across regions and modalities, and a radiomics–clinical nomogram was established using independent clinical predictors. Results: Peritumoral radiomics generally outperformed intratumoral radiomics, with modality-specific optimal margins (CT: 5 mm; PET: 3 mm). Combining intratumoral and peritumoral features further improved discrimination. The multimodal PET/CT fusion model achieved the best performance in external validation (C-index = 0.648). The radiomics–clinical nomogram yielded the highest overall performance (C-index = 0.668) with good calibration and net clinical benefit. Conclusions: Peritumoral radiomics provides critical complementary prognostic information for OS prediction in HNC, and optimal peritumoral extents are modality dependent. The integrated PET/CT-based nomogram demonstrates robust predictive performance and potential clinical utility.