Prediction Of Microvascular Invasion In Solitary Hepatocellular Carcinoma ? 5cm With CT-based Radiomics

Objective: To investigate the value of preoperative prediction of microvascular invasion of ? 5 cm isolated liver cancer based on CT radiomics.It is important to choose a treatment plan for clinical treatment.Methods: A retrospective analysis of 185 patients with partial hepatectomy and pathologically confirmed hepatocellular carcinoma.The clinical information(sex,age,hepatitis,AFP)and preoperative CT arterial phase images were collected.3D-Slice was used to delineate the tumor margin at each layer to form the 3D-ROI of the tumor,and 1351 radiomics features were extracted based on the 3D image.All patients were randomized to a training group(n=124)and a validation group(n=61)in a 2:1 ratio.In the training group,the LASSO feature selection algorithm was used to reduce the dimension of the training set,and the most relevant radiomics features were selected.The selected features were linearly fitted according to the weight of each coefficient,and the image score of each patient was calculated.(Rad-score).The diagnostic performance of radiomics was verified in the validation group,and the radiomics and MVI-related imaging features(TTPVI,RVI)were compared using the Delong test.Results: Of the 185 patients,63 had microvascular invasion and 122 had no microvascular invasion.The features of 1531 image groups werescreened,and 10 features were finally selected.The Rad-score calculation formula is obtained according to the weighting coefficient corresponding to the feature,and the RS score of each patient is calculated.This imaging ensemble model showed better correction and recognition ability in the training and validation groups(AUC: 0.72 [95% CI: 0.58,0.86] and 0.74[95% CI: 0.66,0.83]).The clinical application value of radiomics was confirmed by the receiver working curve analysis,and the predictive performance was significantly higher than the imaging features and the difference was statistically significant(P<0.05).Conclusion: Radiomics has certain value in predicting microvascular invasion of ?5cm isolated hepatocellular carcinoma,and its predictive ability is higher than imaging features.