CT-based intratumoral, peritumoral radiomics and clinical features: a combined model for perineural invasion prediction in PDAC.

PURPOSE: To develop and validate a general radiomics nomogram capable of identifying perineural invasion (PNI) status in pancreatic ductal adenocarcinoma (PDAC) patients.

METHODS

A total of 175 pancreatic cancer patients were retrospectively enrolled in this study. Patients were randomly divided into a training cohort (n = 123) and a test cohort (n = 52) at a ratio of 7:3. Senior physicians manually delineated the intratumoral region of interest (ROI), and the peritumoral ROI was obtained by expanding 3 mm outward from the intratumoral ROI. After extracting and selecting radiomics features, the ExtraTrees algorithm was used to construct intratumoral, peritumoral, and intratumoral + peritumoral radiomics models, respectively. The optimal radiomics model was selected to construct a combined model with clinical characteristics. Receiver operating characteristic (ROC) curves and decision curve analysis (DCA) were used to evaluate the predictive performance of the models.

RESULTS

Multivariate analysis showed that carbohydrate antigen 199 (CA199) (p < 0.05) and vascular invasion (p < 0.05) were associated with an increased risk of PNI. In the training cohort, the area under the curve (AUC), sensitivity, and specificity of the combined model were 0.855, 76.8%, and 80.5%, respectively; in the test cohort, they were 0.844, 76.5%, and 77.8%, respectively. The performance of the combined model was superior to that of the clinical model or radiomics model alone.

CONCLUSIONS

The combined predictive model integrating intratumoral + peritumoral radiomics features based on contrast-enhanced computed tomography (CE-CT) with clinical characteristics can effectively predict PNI in pancreatic cancer.