Predicting Microvascular Invasion Of Hepatocellular Carcinoma Based On Non-contrast-enhanced MRI Radiomics Signature

Objective:Magnetic resonance imaging(MRI)-based radiomics signatures was conducted to predict microvascular invasion(MVI)of hepatocellular carcinoma(HCC)preoperatively.Materials and Methods: 1.General information:129 cases of primary HCC patients who had undergone MRI examination on 3.0T MRI were recruited with MVI confirmed by surgical pathology from 2013.1.1 to 2019.3.20.Among them,104 patients were male and 25 patients were female;77 were MVI positive and 55 were MVI negative.Inclusion criteria include:(1)Resectable solitary and primary HCC confirmed by postoperative pathology;(2)Postoperative pathology clearly indicated whether there was MVI.;(3)Patients with HCC had not received any treatment before operation,such as radiotherapy,chemotherapy and biopsy,etc;(4)All patients had undergone a Siemens verio 3.0 Tesla magnetic resonance imaging(MRI)scan before surgery within two weeks;(5)MRI imaging data and sequence are complete and there are no artifacts.2.Image Processing:All patient MRI images were uploaded to the Radcloud platform 2.1V [Hui Ying Medical Technology(Beijing)Co.,Ltd].The tumor tissue was segmented manually layer by layer.Radiomics features were extracted from fat-suppressed T2-weighted(T2WI-FS)imaging and apparent diffusion coefficient(ADC)map.We used the Variance Threshold,Select KBest,and Least absolute shrinkage and selection operator(LASSO)algorithms in order to perform dimensionality reduction.Then random forests(RF),k-nearest neighbor(KNN),extreme gradient boosting(XGBoost),logistic regression(LR),decision tree(DT)and support vector machine(SVM)algorithm were trained to separate the HCC with MVI positive and with MVI negative.The performance of each model built by the classifier was evaluated by AUC and accuracy.Results:1.The radiomics features were significantly associated with MVI.Quantitative imaging features(n = 1409)were extracted from T2WI-FS and ADC map respectively.Finally,12 features of T2WI-FS and 8 features of ADC were selected to construct the radiomics model separately.These radiomics include: Firstorder(Median、Kurtosis、Energy、Skewness);second-order and high-order texture features(Large Area Low Gray Level Emphasis、Busyness、Large Dependence High Gray Level Emphasis、Small Area High Gray Level Emphasis、Gray Level Variance、Long Run Emphasis、10Percentile、Large Area Emphasis、Large Area High Gray Level Emphasis、Low Gray Level Emphasis、Sum Entropy).2.The model that used SVM and LR classification method achieved the best performance among the six methods,with AUC values of 0.869、0.801 and accuracy of 0.78、0.81.95% CI were 0.70-0.98 and 0.62-0.98.Delong validation was used to compare the areas under the curve(AUCs)in the two models,with a p value of 0.025 indicating the results was statistically significant.Conclusion:1.Good accuracy and AUC could be obtained using only 12 radiomic features of T2WI-FS.2.The model built by SVM classification based on radiomics features from T2WI-FS provides a new auxiliary method for preoperative non-invasive prediction of microvascular invasion of hepatocellular carcinoma.