Development of a CT radiomics nomogram for preoperative prediction of differentiation grading in pancreatic ductal adenocarcinoma: a two-center retrospective study.

aimTo investigate the value of the nomogram based on preoperative enhanced computed tomography (CT) radiomics in predicting the differentiation grade of pancreatic ductal adenocarcinoma (PDAC). MATERIALS AND

methodsA total of 100 patients (66 in the training set and 34 in the validation set) with pathologically confirmed PDAC were derived from two centers. The region of interest (ROI) based on both the arterial phase and the venous phase of the preoperative enhanced CT was automatically drawn by automated computer segmentation algorithm. Subsequently, the automated segmentation results were manually reviewed and corrected by two radiologists using ITK-SNAP. After image resampling and gray-level discretization, 1231 radiomics features were extracted. Feature selection involved stability filtering using the intraclass correlation coefficient (ICC), retaining features with ICC > 0.80. redundancy reduction (Spearman |r| > 0.80), and dimensionality reduction via the least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation. Clinical predictors were identified through univariable and multivariable logistic regression. Three models were developed: a clinical model, a radiomics model (based on the Rad-score from selected features), and a nomogram that integrated both. The area under the receiver operating characteristic (ROC) curves (AUC) and decision curve analysis (DCA) were applied to evaluate the diagnostic efficacy and clinical applicability of the three models. Calibration curves were used to analyze the accuracy of the nomogram.

resultsThe AUC for the nomogram model was significantly higher than that of both the clinical model and the radiomics model in both the training set (0.886, 0.773, and 0.836, respectively) and the validation set (0.842, 0.721, and 0.806, respectively), with all differences being statistically significant (DeLong test, all P < 0.05). The calibration curves demonstrated good consistency between the predicted and the actual probabilities for the nomogram model in both the training and validation sets. DCA confirmed that the nomogram provided the highest clinical net benefit across a wide range of threshold probabilities.

conclusionThe CT-based radiomics nomogram, which integrates clinical risk factors and radiomics, shows strong potential for the preoperative prediction of PDAC differentiation grade. It may serve as a valuable non-invasive tool to aid in individualized clinical decision-making, potentially contributing to optimized patient management.