Predicting overall survival in pancreatic ductal adenocarcinoma using 18F-FDG PET/CT radiomics: a two-center study

Accurate prognostic prediction in pancreatic cancer is of paramount clinical importance for patient management. However, a reliable and non-invasive method for preoperatively predicting overall survival (OS) remains a significant unmet need. We aim to evaluate the utility of pre-operative 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography (PET)/ computed tomography (CT)-derived signatures in predicting OS in patients with pancreatic ductal adenocarcinoma (PDAC). This study included 109 patients (70 males, 39 females). Over a median follow-up period of 28 months (range, 1-64 months), 68 patients (62.4%) died (median OS, 15 months). As stated above, a series of models was established using selected radiomics features and clinical factors. In the test set, the C-index of the clinical, CT-based, PET-based, PET/CT-based, and integrated model were 0.586, 0.724, 0.730, 0.730, and 0.743, respectively. The PET-based model (P = 0.027) significantly outperformed the clinical model, unlike the CT-based model (P = 0.057); further feature integration did not improve performance. In this study, 18F-FDG PET/CT radiomics features showed promise in predicting OS in patients with PDAC, suggesting potential as a tool for personalized management and warranting large-scale studies to confirm its applicability in clinical practice.