Preoperative Discrimination Of Angiomyolipoma From Hepatocellular Carcinoma With MRI-based Radiomics

Background:Angiomyolipoma is a benign mesenchymal tumor that usually presents as a single liver lesion in the background of non-cirrhotic liver.Typical hepatic AML is easy to diagnose because it contains significant fat content,but some have no or only a small amount of mature fat,which is called fat-poor angiomyolipoma(fpAML).It is usually characterized by a single rich blood-supplying solid mass,and the accuracy of preoperative imaging diagnosis is only 20%-52%.It is most easily misdiagnosed as hepatocellular carcinoma(HCC),especially without the background of cirrhosis and negative alpha-fetoprotein hepatocellular carcinoma(n-HCC).The treatment and prognosis of fp-AML and n-HCC are completely different.Therefore,it is of great clinical significance to accurately distinguish between fp-AML and n-HCC before surgery.Conventional imaging has limited value in the differentiation of fpAML and n-HCC.Radiomics can extract massive and quantitative features from images,and mine information that is difficult to be recognized by the naked eye of radiologists.It has great application prospect in tumor differential diagnosis and prognosis prediction.However,there are few reports on the differential between fpAML and n-HCC.Purpose:To explore the value of radiomics model based on magnetic resonance imaging(MRI)in accurate differentiation fp-AML from n-HCC preoperatively in the background of non-cirrhotic liver.Methods:121 patients from the Zhongshan Hospital affiliated to Fudan University,Guangdong Provincial People’s Hospital and Sun Yat-sen University Cancer Center with fp-AML or n-HCC confirmed by surgical pathology from October 2010 to July 2020 were retrospectively analyzed.Among them,41 were patients with fp-AML and 80 were patients with n-HCC.The patients from Zhongshan Hospital affiliated to Fudan University were divided into the training group(fp-AML=25,nHCC=50)and internal test group(fp-AML=6,n-HCC=12)according to entry time,and the patients of other 2 hospitals were divided into external test group(fp-AML=10,nHCC=18).The radiomics features extraction was based on preoperative triple-phase dynamic contrast-enhanced MRI images,respectively.In the training group,the Joint Mutual Information Maximisation(JMIM)method was used to extract the highly associated feature subset and build the radiomics model of differentiation fp-AML from n-HCC.The diagnostic performance of the radiomics model was assessed using the receiver operating characteristic(ROC)curve,and was compared with that of two radiologists with different seniority levels(2 years and 10 years of imaging diagnostic experience).Results:The final established radiomics model showed good performance in differentiation fp-AML from n-HCC in the background of non-cirrhotic liver.In the internal test group,the area under the curve(AUC)of the radiomics model was 0.819(with an accuracy of 72.2%),outperforming than radiologist 2 with 2 years of diagnostic experience(AUC=0.375,P=0.004)and radiologist 1 with 10 years of diagnostic experience(AUC=0.542,P=0.029).In the external test group,the AUC of the radiomics model was 0.772(with an accuracy of 71.4%),which was better than that of radiologist 2(AUC=0.400,P=0.008)and comparable to that of radiologist 1(AUC=0.661,P=0.442).Conclusion:The radiomics model based on triple-phase dynamic contrastenhanced MRI can accurately identify fp-AML and n-HCC in non-cirrhotic liver before surgery,which has potential clinical application value.