Back to all papers

<sup>18</sup>F-FDG PET Radiomic Analysis to Predict Occult Liver Metastases of Pancreatic Ductal Adenocarcinoma.

October 31, 2025pubmed logopapers

Authors

Chen J,Zhang Z,Jin Z,Ma P,Jiang Z,Lu C,Zhu Q,Mou Y,Jin W

Affiliations (3)

  • Department of Gastrointestinal & Pancreatic Surgery, Zhejiang Provincial People's Hospital, Hangzhou 310014, China (J.C., Z.Z., Z.J., P.M., Z.J., C.L., Q.Z., Y.M., W.J.); Postgraduate Training Base Alliance of Wenzhou Medical University (Zhejiang Provincial People's Hospital), Hangzhou 310014, China (J.C., Z.Z., P.M.).
  • Department of Gastrointestinal & Pancreatic Surgery, Zhejiang Provincial People's Hospital, Hangzhou 310014, China (J.C., Z.Z., Z.J., P.M., Z.J., C.L., Q.Z., Y.M., W.J.).
  • Department of Gastrointestinal & Pancreatic Surgery, Zhejiang Provincial People's Hospital, Hangzhou 310014, China (J.C., Z.Z., Z.J., P.M., Z.J., C.L., Q.Z., Y.M., W.J.). Electronic address: [email protected].

Abstract

To develop and validate a preoperative predictive model for occult liver metastases (OLM) in pancreatic ductal adenocarcinoma (PDAC) using fluorine-18 fluorodeoxyglucose positron emission tomography (<sup>18</sup>F-FDG PET) radiomics. This retrospective study included 117 patients with PDAC who underwent preoperative <sup>18</sup>F-FDG PET/CT and surgical resection. OLM was defined as liver metastases detected during surgery or within 6 months postoperatively. A fully automated pancreas segmentation strategy was employed, and radiomic features were extracted from PET images. Three machine learning models (logistic regression, multilayer perceptron, and adaptive boosting) were developed and compared to a clinical model incorporating jaundice, metabolic tumor diameter, and maximum standardized uptake value. A fusion model integrating PET radiomic features with clinical variables was subsequently constructed. Model performance was evaluated using receiver operating characteristic curves and decision curve analysis. Among the 117 patients, 15.4% (n=18) had OLM. The logistic regression radiomics model demonstrated favorable predictive performance (area under the curve [AUC]: 0.936 in the testing cohort) compared to a clinical model based on conventional parameters (AUC: 0.755, P<0.001). Subgroup analyses confirmed robustness across different jaundice statuses, tumor locations, and carbohydrate antigen 19-9 levels. The fusion model that integrates radiomic and clinical features provides a comprehensive tool for preoperative risk stratification, with the potential to guide personalized treatment strategies. In this exploratory study, the <sup>18</sup>F-FDG PET radiomics model demonstrates promising predictive performance for OLM in PDAC, outperforming conventional clinical parameters. It shows potential as a valuable tool for preoperative risk stratification and may help inform personalized treatment planning.

Topics

Journal Article

Ready to Sharpen Your Edge?

Join hundreds of your peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.