Predictive value of 18F-FDG PET/CT delta radiomics model for prognosis after neoadjuvant therapy for locally advanced pancreatic cancer.

Purpose: To develop a delta radiomics model based on 18F-FDG PET/CT to predict prognosis after NAT for LAPC. Materials and methods: This study retrospectively collected data from 33 patients with LAPC admitted to our hospital from August 2019 to July 2024. The volume of the regions of interests (VOIs) of LAPC lesions was outlined on 18F-FDG PET/CT images before and after neoadjuvant therapy. Radiomics features were extracted based on the VOIs, and the subtraction of the two radiomics features was the delta radiomics feature. Feature screening was performed using t-test and LASSO-COX regression to build a prediction model. ROC curves were used to assess the predictive ability of the prediction model for progression-free survival (PFS). Calibration curve and decision curve analysis (DCA) were used to assess the accuracy and net benefit of the prediction model, respectively. Results: Univariate Cox regression analysis demonstrated that neither clinical parameters nor 18F-FDG PET/CT metabolic parameters were independent factors in predicting PFS in LAPC. A predictive model was constructed based on two delta radiomics features screened by LASSO-COX regression. The delta radiomics model demonstrated a high predictive ability for PFS with a C-index of 0.78 (95%CI: 0.68-0.88). High-risk patients identified by delta features showed markedly shorter PFS (HR:5.357, P = 0.006). DCA indicated favorable clinical utility for the delta radiomics model. Conclusion: The delta radiomics model based on 18F-FDG PET/CT, provides a non-invasive tool to predict prognosis in LAPC after NAT, which may facilitate personalized treatment strategies.