A radiomics-based model for predicting lymph nodes metastasis of pancreatic ductal adenocarcinoma: a multicenter study

Purpose To develop a radiomics model to predict lymph nodes metastasis (LNM) in patients with pancreatic ductal adenocarcinoma (PDAC) and assess its value for clinical management. Methods Patients with pathologically confirmed PDAC from four centers were retrospectively enrolled and split into four cohorts: training (n = 192), validation (n = 82), testing (n = 100), and clinical utilization (n = 163). A radiomics model was constructed based on contrast-enhanced CT (CECT) to predict LNM, and its performance was evaluated using the areas under the curve (AUC). Kaplan-Meier analysis was used to assess the prognostic and therapeutic decision-assisting value of the radiomics model. Results A total of 437 patients (mean age: 63.1 years +/- 9.2 standard deviation; 253 men) were included. The radiomics model outperformed other models with AUCs of 0.84, 0.82, and 0.78 in the training, validation, and testing cohorts (all p < 0.05), respectively. LNM predicted by the radiomics model was significantly associated with overall survival (p < 0.001). Kaplan-Meier analysis revealed that patients with a higher risk of LNM also had worse outcomes (all p < 0.05). Additionally, among the high-risk subgroup identified by the radiomics model in the clinical utilization cohort, patients who underwent dissection of >= 15 lymph nodes exhibited better overall survival compared to those with fewer lymph nodes dissected (p = 0.002). Conclusion The radiomics model we constructed demonstrated impressive performance in predicting LNM and prognosis, suggesting its potential for optimizing the clinical management of PDAC. Critical relevance statement This radiomics model can predict the risk of lymph nodes metastasis and prognosis of patients in pancreatic ductal adenocarcinoma and has potential value in selecting patients who can benefit from different extents of lymph nodes dissection.