Objective Development and validation of a nomogram model for preoperative prediction of microvascular in-vasion(MVI)in colorectal cancer liver metastases(CRLM)based on Gd-EOB-DTPA enhanced MRI hepatobiliary phase radiomics features.Methods A total of 63 patients who underwent Gd-EOB-DTPA enhanced MRI and were pathologically diagnosed with CRLM were retrospectively and consecutively collected,of which 22 were MVI-positive and 41 were MVI-negative.Univariate and multivariate binary logistic regression analysis were applied to analyze the clinical indicators such as gender,age,tumor indexes and MRI features to screen out the clinical and MRI features that were statistically significant in predicting the MVI of patients with CRLM and to establish a clinical model.Radiomics features were extracted and screened from the hepatobiliary phase of Gd-EOB-DTPA enhanced MRI.Finally,three classifiers,logistic regression(LR),linear discriminant analysis(LDA)and support vector machine(SVM),were used to construct the radiomics models and using fivefold cross validation.In addition,a nomogram model combining the clinical model and the best radiomics model was constructed.Efficacy was evaluated by receiver operating characteristic(ROC)curves,calibration curves and decision curve analysis(DCA).Results The univariate and multivariate results showed that T2WI signal inhomogeneity and peri-tumor low signal in the hepatobiliary stage were independent predictors of MVI in CRLM patients.The radiomics model constructed by the final LR classifier had relatively high predictive efficacy in both the training and test groups.The nomo-gram model constructed by the clinical model combined with the radiomics model had the best predictive efficacy,with AUCs of 0.970(95％CI:0.911-1.000)and 0.917(95％CI:0.798-1.000)in the training and test groups,respec-tively.The calibration curves demonstrated the goodness of fit of the nomogram model,and the DCA indicated that the no-mogram model had a high net clinical benefit rate in predicting MVI in patients with CRLM.Conclusion The nomogram model based on Gd-EOB-DTPA enhanced MRI hepatobiliary phase radiomics combined with MRI features(T2WI signal in-homogeneity and peri-tumor low signal in the hepatobiliary stage)can be used as an auxiliary tool for preoperative detection of MVI in patients with CRLM.
Microvascular invasionColorectal cancer liver metastasesRadiomicsGd-EOB-DTPA
NETL
NSTL
万方数据
