CT texture analysis for the presurgical prediction of superior mesenteric–portal vein invasion in pancreatic ductal adenocarcinoma: comparison with CT imaging features

AIM To investigate the value of computed tomography (CT) texture analysis (TA) and imaging features for evaluating suspected surgical superior mesenteric–portal vein (SMPV) invasion in patients with pancreatic ductal adenocarcinoma (PDAC). MATERIALS AND METHODS Fifty-four patients with PDAC in the pancreatic head or uncinate process with suspected SMPV involvement were analysed retrospectively. SMPV invasion status was identified by surgical exploration. For each patient, 396 texture features were extracted on pretreatment CT. Non-parametric tests and minimum redundancy maximum relevance were used for feature selection. A CTTA model was constructed using multivariate logistic regression, and the area under the receiver operating characteristic (AUROC) of the model was calculated. Two reviewers evaluated qualitative imaging features independently for SMPV invasion and interobserver agreement was investigated. The diagnostic performance of the imaging features and the CTTA model for SMPV invasion was compared using the McNemar test. RESULTS Of the 54 patients with PDAC, SMPV invasion was detected in 23 (42.6%). The CTTA model yielded an AUROC of 0.88 (95% confidence interval, 0.76–0.97) and achieved significantly higher specificity (0.90) than the two reviewers (0.61 and 0.65; p=0.027 and 0.043). Interobserver agreement was moderate between the two reviewers (κ = 0.517). Of the 13 cases with disagreement between the two reviewers, 11 cases were predicted accurately by the CTTA model. CONCLUSION CTTA can predict suspected SMPV invasion in PDAC and may be a beneficial addition for qualitative imaging evaluation.