Objective To investigate the predictive value of computed tomography(CT) based radiomics model for the prognosis of patients with pancreatic ductal adenocarcinoma(PDAC).

Methods The retrospective cohort study was conducted. The clinicopathological data of 206 PDAC patients who were admitted to Zhongshan Hospital of Fudan University from August 2018 to December 2020 were collected. There were 115 males and 91 females, aged (64±9)years. All 206 pati-ents underwent enhanced CT examination. Based on radom number table, the 206 patients were randomly divided into a training set of 165 cases and a validation set of 41 cases with a ratio of 4∶1. The training set was used to construct the prediction model, and the test set was used to validate the performance of the prediction model. Observation indicators: (1) follow‑up; (2) analysis of prognostic factors of PDAC patients in the training set; (3) construction and evaluation of prediction model for prognosis of PDAC patients. Comparison of measurement data with normal distribution between groups was conducted using the t test. Comparison of measurement data with skewed distribution between groups was conducted using the Wilcoxon W test. Comparison of count data between groups was conducted using the chi‑square test or corrected chi‑square test. The Kaplan‑Meier method was used to calculate the survival rate and Log‑rank test was used for survival analysis. Univariate and multivariate analyses were conducted using the COX regression model. The PyCharm software was used for the least absolute shrinkage and selection operator method (LASSO)-COX regression analysis. The receiver operating characteristic curve was plotted to evaluate the performance of radiomics model.

Results (1)Follow‑up. Of the 206 patients,205 cases were followed up for 17.1(range, 12.0‒40.1)months. The postoperative 1‑, 2‑, 3‑year survival rates were 80.10%, 29.61% and 4.85%. (2) Analysis of prognostic factors for PDAC patients in the training dataset. Results of multivariate analysis showed that pathological N stage was an independent influencing factor for prognosis of PDAC patients in the training set (hazard ratio=1.476, 95% confidence interval as 1.054‒2.067, P<0.05). (3) Construction and evaluation of prediction model for prognosis of PDAC patients. A total of 1 595 radiomics features were finally extracted from the 206 patients. By intra‑group feature selection and dimensionality reduction using LASSO‑COX regression model, 10 radiomics features were obtained. Combined with 10 radiomics features and 11 clinical features, using the LASSO‑COX regression analysis, 15 features were finally extracted to construct the CT based radiomics model for predicting prognosis of PDAC. The areas under receiver operating characteristic curve of the prediction model in predicting 2‑year and 3‑year overall survival rates of PDAC patients in the training set were 0.834 (95% confidence interval as 0.777‒0.891) and 0.883 (95% confidence interval as 0.834‒0.932), respectively. The area under curve of the prediction model for patients in the validation set was 0.606 (95% confidence interval as 0.456‒0.756) and 0.625 (95% confidence interval as 0.477‒0.773).

Conclusion The prediction model constructed on CT based radiomics features and clinical features for predicting the prognosis of PDAC patients shows a promising prediction efficiency.