Evaluation of microvascular invasion in hepatocellular carcinoma based on CT-enhanced portal venous phase radiomics
10.3969/j.issn.1002-1671.2025.07.015
- VernacularTitle:基于CT增强门静脉期影像组学评价肝细胞癌微血管侵犯
- Author:
Mengchen YANG
1
;
Tianmin ZHOU
1
;
Yanming ZHANG
1
;
Shangyu YANG
1
;
Haiyang LIU
1
Author Information
1. 商洛市中心医院医学影像科,陕西 商洛 726000
- Publication Type:Journal Article
- Keywords:
computed tomography;
radiomics;
machine learn-ing;
hepatocellular carcinoma;
microvascular invasion
- From:
Journal of Practical Radiology
2025;41(7):1148-1152
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the value of CT-enhanced portal venous phase radiomics combined with machine learning algo-rithms in assessing microvascular invasion(MVI)in hepatocellular carcinoma(HCC).Methods A retrospective analysis was con-ducted on imaging and clinical data of 132 HCC patients.The patients were randomly divided into training set and test set at a 7︰3 ratio.Independent influencing factors for predicting MVI status in HCC patients were identified through univariate and multivariate logistic regression analyses.Radiomics features were selected using the least absolute shrinkage and selection operator(LASSO)algorithm,and a Radiomics score(Radscore)was calculated to construct the final combined model.Four machine learning algorithms including-logistic regression(LR),naive Bayes(NB),support vector machine(SVM),and K-nearest neighbor(KNN)were applied to evaluate the model.The performance of each machine model was assessed using the receiver operating characteristic(ROC)curves and the area under the curve(AUC).Results Univariate and multivariate logistic regression analyses revealed that venous phase CT values were independentinfluencing factors,and six radiomics features were ultimately selected.After the Radscore was calculated,a combined model was constructed using Radscore and venous phase CT values.Machine learning algorithms showed that the combined model achieved the following AUC in the training set:0.895[95%confidence interval(CI)0.821-0.965]for LR,0.892(95%CI 0.831-0.963)for NB,0.644(95%CI 0.532-0.765)for SVM,and 0.855(95%CI 0.783-0.947)for KNN.In the test set,the respective AUC were 0.845(95%CI 0.712-0.961),0.840(95%CI 0.723-0.964),0.492(95%CI 0.311-0.687),and 0.716(95%CI 0.566-0.871).Conclusion Radiomics based on CT-enhanced portal venous phase combined with machine learning algorithms demonstrates high efficiency in preoperative evaluation of MVI in HCC,with the LR model showing the best performance.
