Abstract
Pancreatic cancer (PaCa) is the seventh leading cause of cancer deaths globally, with limited detection and treatment. Pancreatoduodenectomy (PD) is the primary surgical intervention for resectable pancreatic head cancers (PaHCa), but its complexity necessitates prognostic tools. This study evaluates the arterial and venous phase radiomic features role from preoperative CT scans in predicting survival for PaHCa patients undergoing pylorus-preserving pancreatoduodenectomy (PPPD). A retrospective analysis was conducted on 42 PaHCa patients (mean age 63.3 ± 10 years; 20 males, 22 females) who underwent PPPD between 2010 and 2017. Radiomic features were extracted from arterial and venous phase CT images, and a gradient boosting survival model was applied for survival prediction using cross-validation. Ethical approval (Approval number: EK028/19, date: 03.05.2019) was granted, and informed consent was waived due to the retrospective nature of the study and all experiments were performed in accordance with relevant guidelines and regulations. No identifying information or images are included in this manuscript. Survival analysis revealed no significant differences when using arterial (p = 0.161) or venous (p = 0.668) phase features alone. However, combining arterial and venous phase features significantly improved survival prediction (p = 0.007). Key predictive features included "Shape: Sphericity" and "Gray Level Size-Zone Non-Uniformity (GLSZM)". Combining arterial and venous phase radiomic features enhances survival prediction in PPPD-treated PaHCa patients, highlighting the potential of multi-phase CT radiomics for personalized treatment strategies. Radiomics-based survival prediction of PaHCa prior to patients undergoing PPPD may guide clinical decision-making and improve personalized treatment planning.