Exploring The Value Of Preoperative Predictability Microvascular Invasion Of Hepatocellular Carcinoma By MRI Radiomic

Purpose:The purpose of this study was to explore the value of preoperative predict HCC MVI based on radiomic of magnetic resonance imaging(MRI),that can provide reference for clinical surgical selection.Methods:196 patients with HCC confirmed by pathology in The Hunan Provincial People’s Hospital from September 2014 to September 2018,were enrolled in our study retrospectively.According to the postoperative pathology,the patients were divided into positive MVI[MVI(+)]group including 85 patients,and negative MVI[MVI(-)]group including 111patients.All patients underwent normal MRI and dynamic contrast-enhanc-ed MRI(DCE-MRI)before surgery.Radiomic features were extracted fro-m HCC of the DCE-MRI Late arterial phase(LAP)original images.The patients were randomly divided into a training group including 156patients and a verification group including 40 patients by 4:1.In the training group,MVI-related radiomic features were selected by intraclass correlation coefficient(ICC),the least absolute shrinkage and selection operator(LASSO)was used to reduce the dimension,and the most relevant radiomic features were selected.A logistic regression(Binary Logistic)is used to set up radiomic model.Assessment subjective image features of HCC on DCE-MRI LAP original images,collect clinical data[including age,gender,alpha-fetoprotein(AFP),cause of liver disease(whether it is hepatitis B),Neutrophil to Lymphocyte Ratio(NLR)]by consulting the electronic medical record system,whether data have statistically significant differences between the MVI(+)group and the MVI(-)group is obtained by T test and independent sampleχ~2 test,and then establish an image-clinical model(shorten as clinical model)by logistic regression.Finally,a combined diagnostic model(shorten as combined model)containing information in the radiomic model and the clinical model was established.The MVI situation of the training group is predicted separately by the established three models,and then the verification group is internally verified.Three models were used to predict the efficacy of HCC MVI using the area under the receiver operating characteristic curve(AUC),at last but not at least,the Delong test was used to compare whether the difference in efficacy between the models was statistically significant.Results:(1)The radiomic model consists of six radiomic features,and which is helpful for predicting HCC MVI status(AUC:training group0.717,validation group 0.687;P<0.05).(2)The differences between MVI(+)group and MVI(-)group of the clinical information(as AFP)and the subjective image characteristics[as background liver tissue,nonrim arterial phase hyper enhancement(APHE),nonperipheral washout,corona enhancement,hemorrhage]have statistically significant(P<0.05).The difference between MVI(+)group and MVI(-)group of age,gender,cause of liver disease,NLR,tumor size have no statistically significant(P≥0.05).And the clinical model has positive performance in predicting HCC MVI(AUC:training group 0.742,validation group 0.757;P<0.05).(3)In three models,the combined model has the best predictive power(AUC:training group 0.817,validation group 0.799;P<0.05).Conclusion:The radiomic model based on MRI images can predict the status of MVI in HCC before surgery,which has the same predictive performance as clinical models;and in three models,the combined model has the best predictive power.This provides a reference for the clinical use of non-invasive techniques to select surgical treatment.