Objective To observe the value of enhanced CT radiomics combined with clinical indicators for predicting of extramural venous invasion(EMVI+)of colorectal cancer.Methods Data of 131 patients with colorectal cancer proved by surgery pathology were retrospectively analyzed.The patients were divided into training set(n=92,including 44 cases with EMVI+ and 48 with EMVI-)and test set(n=39,including 23 cases with EMVI+ and 16 with EMVI-)at the ratio of 7∶3.The best radiomics features were extracted based on preoperative portal-venous phase CT to construct a radiomics model.Univariate and multivariate logistic regression analysis were used to analyze the clinical,CT and pathological data of the training set,and the independent predictors of colorectal cancer EMVI were screened to build a clinical model.Finally a combined model was established based on radiomics and clinical model.Receiver operating characteristic curve was drawn,and the area under the curve(AUC)was calculated to evaluate the efficacy of each model for predicting EMVI+ in colorectal cancer.Calibration curve and decision curve analysis were used to evaluate the calibration degree and clinical practicability of the models.Results Four best radiomics features were selected to construct the radiomics model.Carbohydrate antigen(CA)19-9 and CA 72-4 were both independent predictors of EMVI+ for colorectal cancer(OR=1.033,1.285,both P＜0.05).The AUC of combined model(AUC=0.908)for predicting EMVI+ of colorectal cancer in training set was higher than that of radiomics and clinical models(AUC=0.825,0.770,P=0.017,0.003).In test set,the AUC of radiomics,clinical and combined models was 0.751,0.632 and 0.799,respectively,not being statistical different between each pair(all P＞0.05).The radiomics model and combined model both had good calibration degree.Taken ＞0.1 in training set and ＞0.12 in test set as the thresholds,the clinical net benefit of combined model was higher.Conclusion Enhanced CT radiomics combined with clinical indicators could effectively predict EMVI+of colorectal cancer.
维普
万方数据