Pancreatic Adenocarcinoma : Quantitative CT Features Are Correlated With Fibrous Stromal Fraction And Help Predict Outcome After Resection

Objectives: Pancreatic ductal adenocarcinoma(PDAC)is one of the most lethal malignant neoplasms across the world and the fourth most common cause of cancer-related death among U.S.Here,we identify quantitative imaging features of contrast-enhanced computed tomography(CE-CT)that may be prognostically favorable after resection of smaller(?30 mm)PDACs located within the head.Methods: This retrospective study included two independent cohorts(discovery cohort,n=212;test cohort,n=100)of patients who underwent surgical resection of head PDACs?30 mm and preoperative CE-CT studies.Preoperative CT images were first imported into the open-source 3D Slicer v4.10 application for precise image registration in each patient.Next,volumes of interest(VOI)in tumor and adjacent pancreas were segmented manually by one radiologist.Mask files were then saved and imported into Matlab to acquire CT attenuations at each pixel and to calculate volumetric mean attenuations of tumor and parenchyma at all registered phases.We examined tumor and surrounding parenchymal attenuation differences(deltas),as well as tumor attenuation changes across phases(ratios).Qualitative semantic features of PDACs were evaluated by two radiologists.Clinicopathologic and imaging features for predicting disease-free survival(DFS)and overall survival(OS)were analyzed via multivariate Lasso-penalized Cox proportional-hazards models.Survival rates were derived by Kaplan-Meier method and compared using log-rank test.Results : Imaging features achieving C-indices of 0.766(discovery cohort)and 0.739(test cohort)for DFS,0.790(discovery cohort)and 0.772(test cohort)for OS estimates through incorporation of clinicopathologic features.The most decisive imaging feature was attenuation differences between tumor and surrounding pancreas at pancreatic phase(DFS: HR=2.122;OS: HR=2.375;both p<0.001).Compared with inconspicuous(<28HU)tumors,conspicuous(? 28 HU)tumors correlated significantly with more aggressive histologic grades(p=0.014)and less extensive tumor stromal fractions(p<0.001).Patients with inconspicuous tumors and ?20 mm tumors experienced the most favorable clinical outcomes(DFS,36 months;OS,42 months),whereas those with conspicuous tumors fared poorly,regardless of tumor size(DFS,12 months;OS,19 months).Conclusions: Quantifiable CT imaging features reflect heterogeneous stromal fractions and histologic grades of PDAC at head locations that help stratify patients with disparate clinical outcomes.