An MRI-based Radiomics Nomogram For Preoperative Prediction Of Microvascular Invasion In Hepatocellular Carcinoma

Objective: To develop and validate a Magnetic Resonance Imaging(MRI)-based radiomics nomogram for preoperative prediction Microvascular Invasion(MVI)of hepatocellular carcinoma(HCC),which will help guide the treatment of HCC patients.Methods: The clinical biochemical indexes,radiological and pathological data of 190 patients with HCC who received radical hepatectomy(open or laparoscopic assisted hepatectomy)treatment in department of hepatobiliary surgery(The First Affiliated Hospital of Army Military Medical University)from September 2017 to May 2020 was prospectively collected..According to the ratio of 7:3,190 patients with HCC were randomly divided into training subset(n=130)and test subset(n=60).According to histopathological diagnosis of MVI,training subset were divided into MVI(-)group and MVI(+)group.The clinical biochemical indexes and imaging characteristics related to MVI were analyzed by Logistic regression.Using the images of arterial phase(AP)and hepatobiliary phase(HBP)of Gadolinium ethoxybenzyl-diethylenetriaminepentaacetic acid(Gd-EOB-DTPA)-enhanced MR,the least absolute shrinkage and selection operator(LASSO)algorithm were used to extract and selected radiomics features with low correlation coefficient and high correlation with MVI through the each volume of interest(VOI).The VOI includes the area within the tumor and within 1 cm around the tumor.Logistic regression was used to establish the prediction model in the training subset and the test subset respectively,and the receiver operating characteristic(ROC)curve was constructed to evaluate the prediction model.Results: There were 54 cases in MVI(-)group and 76 cases in MVI(+)group.Logistic regression analysis showed that alpha fetoprotein(AFP)and tumor size was an independent risk factor for MVI.From each VOI of AP and HBP images of Gd-EOB-DTPA-enhanced MR,1748 radiomics features were extracted,and 21 radiomics features((including 10 AP features and 11 HBP features)related to MVI were selected.The radiomics feature model were established in the training subset and the test subset,respectively,and the area under the receiver operating characteristic curve(AUC)of the training subset and the test subset were0.857(95% CI:0.794-0.920)and 0.718(95% CI:0.585-0.851),respectively.The Nomogram prediction model was constructed by combining the radscore,AFP and tumor size,and the AUC of nomogram model was 0.892(95% CI:0.838-0.946)in training subset and 0.717(95% CI:0.585-0.849)in test subset.Conclusions: Nomogram was established to predict the MVI of HCC by combining the radscore,AFP and tumor size,which plays a certain role in guiding the treatment plan of HCC patients.