To explore the preoperative predictive factors for peritoneal metastasis in patients with advanced gastric cancer and construct a nomogram prediction model combining clinical and radiomics features. A total of 342 patients with advanced gastric cancer who were scheduled to undergo surgical treatment were selected and randomly divided into a training set (239 cases) and a validation set (103 cases) at a ratio of 7:3. In the training set, patients were divided into a peritoneal metastasis group and a non-peritoneal metastasis group according to whether peritoneal metastasis was confirmed by postoperative pathology. Preoperative clinical data, laboratory indicators, CT imaging features, and radiomics features of patients were collected. A prediction model was constructed through multivariate logistic regression analysis, and the model’s efficacy was evaluated by receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA). Among the 342 patients with advanced gastric cancer, 96 cases (40.17%) in the training set and 32 cases (31.07%) in the validation set had peritoneal metastasis. Multivariate logistic regression analysis showed that the maximum depth of ascites, the difference in omentum CT value, the longest diameter of the largest peritoneal nodule, tumor volume, tumor texture Gray-level co-occurrence matrix (GLCM) contrast, and CA125 were independent predictive factors for peritoneal metastasis in advanced gastric cancer (all P 0.05. DCA showed that the model had significant clinical net benefit within the threshold probability range of 0.1–0.8. The nomogram prediction model for peritoneal metastasis in advanced gastric cancer based on preoperative clinical-radiomics features has good predictive efficacy and clinical practicability, which can provide a reference for preoperative individualized risk assessment.
Qi et al. (Fri,) studied this question.