Research On Prediction Of Microvascular Invasion In Primary Hepatocellular Carcinoma Based On Ultrasound Radiomics

Primary hepatocellular carcinoma(HCC)is one of the most common malignant tumors in the world,with the second highest mortality rate.Among them,the status of microvascular invasion(MVI)is related to the choice of treatment plan and prognosis evaluation of patients,while the clinical diagnosis of microvascular invasion only depends on postoperative histopathological examination.Therefore,it is of great clinical significance to construct an accurate preoperative microvascular invasion state prediction model for patients with primary hepatocellular carcinoma.Imaging histology can extract non-invasive diagnosis and predict prognosis of tumors by extracting high-throughput information from images for deep mining and analysis.However,existing imaging omics methods mainly focus on the tumor area and ignore the surrounding environment of the tumor.In pathology,the tissue surrounding the tumor is the first area affected by microvascular invasion.Therefore,compared with the tumor area,the imaging features involving the surrounding tissue of the tumor may be shown to be directly related to microvascular invasion.At the same time,ultrasound examination has the characteristics of simple operation,non-invasive and portable,and is the preferred examination method for clinical liver imaging examination.Therefore,this study proposes for the first time to build an ultrasound prediction model of the microvascular invasion state of patients with primary hepatocellular carcinoma based on the peritumoral region,and to build a more comprehensive and accurate auxiliary diagnostic model,which is expected to solve the preoperative diagnosis of microvascular invasion state.The data of this study were from Zhongshan Hospital Affiliated to Shanghai Fudan University.A total of 322 patients with primary hepatocellular carcinoma confirmed by histopathology(including 178 patients with negative microvascular invasion and 102 patients with M1),42 M2 patients)ultrasound images and clinical information,and divided into training test sets according to different instrument models.For the first time,an imaging omics model based on tumor and peritumor was proposed,and for the first time,M1 and M2 classification models with positive microvessel invasion were established to achieve a comprehensive and accurate prediction of the state of microvessel invasion in patients with primary hepatocellular carcinoma.First,extract high-throughput imaging histology features from the tumor area(GTR)and the peritumoral area(PTR)respectively,filter the features through machine learning methods and use random forest(RF),support vector machine(SVM)and LASSO to build a classification model Finally,the area under the ROC curve(AUC)is used to evaluate and compare the classification ability of different models.The above-constructed model was tested in the validation set.The results of the study showed that the imaging histology model based on the peritumoral area can be better identified(AUC=0.710)for the identification of positive microvessel violations(AUC=0.710),and will be based on the tumor area and The model established by the logistic regression of the peritumoral region(AUC=0.726)is superior to the model established by using the tumor alone(AUC=0.708)or the characteristics of the peritumoral region,showing that the peritumoral region has a synergistic effect.At the same time,the study found that if the imaging histology model and the clinical alpha-fetoprotein(AFP)indicators were fused to build a human-machine combination model,the accuracy of predicting the positive of microvessel invasion was further improved(AUC=0.744);for the identification of M1 and M2,The imaging histology model based on the tumor area has good classification ability(AUC=0.806).The significance of this study is to verify the important value of the peritumoral region clinically neglected in the past for the prediction of microvascular invasion status,and for the first time to classify and predict M1 and M2 in positive micro vascular invasion,combined with clinical signs to comprehensively and objectively predict microvascular invasion The status is expected to assist doctors in clinical practice for accurate and non-invasive diagnosis of microvascular invasion in patients with primary hepatocellular carcinoma.