A CT-based Radiomics Model for Prediction of Lymph Node Metastasis in Early Stage Gastric Cancer

Rationale and Objectives To develop and validate a CT-based radiomics model for preoperative prediction of lymph node metastasis (LNM) in early stage gastric cancer (EGC). Materials and Methods Four hundred and sixty-three consecutive EGC patients were enrolled in this retrospective study. Radiomics features were extracted from portal venous phase CT scans. A radiomics signature was built based on highly reproducible features using the least absolute shrinkage and selection operator method. The predictive performance of radiomics signature was tested in the training and testing cohorts. Multivariate logistic regression analysis was conducted to build a radiomics-based model combined radiomics signature and lymph node status according to CT. Performance of the model was determined by its discrimination, calibration, and clinical usefulness. Results The radiomics signature comprised six robust features showed significant association with LNM in both cohorts. A radiomics model that incorporated radiomics signature and CT-reported lymph node status showed good calibration and discrimination in the training cohort (AUC = 0.91) and testing cohort (AUC = 0.89). Decision curve analysis confirmed the clinical utility of this model. Conclusion The CT-based radiomics model showed favorable accuracy for prediction of LNM in EGC and may help to improve clinical decision-making. To develop and validate a CT-based radiomics model for preoperative prediction of lymph node metastasis (LNM) in early stage gastric cancer (EGC). Four hundred and sixty-three consecutive EGC patients were enrolled in this retrospective study. Radiomics features were extracted from portal venous phase CT scans. A radiomics signature was built based on highly reproducible features using the least absolute shrinkage and selection operator method. The predictive performance of radiomics signature was tested in the training and testing cohorts. Multivariate logistic regression analysis was conducted to build a radiomics-based model combined radiomics signature and lymph node status according to CT. Performance of the model was determined by its discrimination, calibration, and clinical usefulness. The radiomics signature comprised six robust features showed significant association with LNM in both cohorts. A radiomics model that incorporated radiomics signature and CT-reported lymph node status showed good calibration and discrimination in the training cohort (AUC = 0.91) and testing cohort (AUC = 0.89). Decision curve analysis confirmed the clinical utility of this model. The CT-based radiomics model showed favorable accuracy for prediction of LNM in EGC and may help to improve clinical decision-making.