| Background and AimsHepatocellular carcinoma(HCC)is a global health problem and is the third leading cause of cancer-related death worldwide.In China,HCC ranks the fourth among common malignancies,and its new cases and deaths account for about 50%of the world,posing a serious threat to human life and health.After clinical interventions such as surgical resection or liver transplantation,5 years of postoperative recurrence rates of HCC are still as high as 70%and 35%.Studies have shown that microvascular invasion(MVI)is the most direct predictive signal of intrahepatic micrometastases and an important source of tumor recurrence in HCC,and its induced intrahepatic dissemination and occult metastasis are a major factor in tumor recurrence after liver resection or liver transplantation.MVI usually refers to the nests of cancer cells seen in the vascular lumen lining with endothelial cells under a microscope.It is common in the portal vein branches of the liver parenchyma around the tumor,followed by hepatic vein branches,and small branches of the hepatic artery,bile duct and lymphatic vessels are rare.Currently,MVI can only be diagnosed by invasive techniques such as histopathology of surgical or biopsy specimens,with a significant lag.If MVI can be accurately predicted before surgery,it will help to formulate treatment strategies in clinic,thereby improving the survival and quality of life of the patients.Magnetic resonance imaging(MRI)has become the main technique for clinical diagnosis and evaluation of HCC due to its good soft tissue resolution,especially the application of Gadolinium ethoxybenzoyl diethylenetriamine pentaacetic acid(Gd-EOB-DTPA).As a hepatocyte-specific contrast agent,it can be uptaken by normal hepatocytes during the hepatobiliary phase.When the normal hepatocyte function is reduced or replaced by tumor cells,the corresponding regional liver tissue shows a low signal due to the dysfunction of the contrast agent in the uptake during the hepatobiliary phase.Therefore,enhancing MRI with Gd-EOB-DTPA can not only obtain the multiphase dynamic enhancement effect of conventional contrast agents,but also evaluate hepatocyte function,which can more accurately evaluate MVI status from many aspects of tumor morphology,hemodynamics and functional imaging.However,the prediction of MVI based on MR imaging features is greatly affected by the image quality,the clinical experience and subjective consciousness of radiologists,and there are differences in the prediction accuracy.With the rise of artificial intelligence in recent years,radiomics plays a crucial role in predicting MVI by extracting quantitative imaging features from medical images.Many studies have shown that radiomics has a good predictive effect,but some researchers believe that the predictive effect of radiomics is not as good as the imaging features,and it is unclear whether the fusion of imaging features and radiomics features can certainly improve the predictive effect of both.Therefore,the purpose of this study was to construct and compare the efficacy of clinicoradiological model including clinical laboratory indicators and radiographic features,radiomics model and clinicoradiological-radiomics hybrid model in predicting MVI based on Gd-EOB-DTPA enhanced MRI examination.Material and methods1.Patient enrollmentA total of 602 HCC patients undergoing curative resection from two institutions were recruited for the study.A total of 501 patients from Institution I(First Affiliated to Army Military Medical University)were randomly divided into a training cohort(351 cases)and a validation cohort(150 cases)in a ratio of 7:3,and then 67 HCC patients were included in institution I as the prospective validation cohort.34 HCC patients from Institution II(The Second Affiliated Hospital of Chongqing Medical University)served as the external validation cohort.2.Image acquisitionAll patients from institution I underwent 3.0T MRI examination,and all patients from institution II underwent 1.5T MRI examination,and obtained T1WI,T2WI,arterial phase,portal venous phase,transition phase,hepatobiliary phase,and diffusion weighted imaging(DWI)images in turn,and then the apparent diffusion coefficient(ADC)values were calculated with a single index function with a b value of 0 and 800s/mm~2.3.Lesions delineationThe T1WI,T2WI,arterial phase,portal venous phase,DWI and ADC images of all enrolled patients were uploaded to the 3D Slicer software,and the tumor volume was manually segmented layer by layer on each sequence,including the entire tumor and intratumoral hemorrhage and necrosis areas,avoiding Peritumoral area of abnormal enhancement.4.Model buildingThe Least absolute shrinkage and selection operator(Lasso)method was used to screen the most important clinicoradiological and radiomic features for preoperative prediction of MVI.The three machine learning algorithms of logistic regression(LR),random forest(RF)and support vector machine(SVM)were used to establish the clinicoradiological,radiomics,and clinicoradiological-radiomics hybrid models.Area under the curve(AUC)of receiver operating characteristic(ROC)curves and Delong’s test were used to compare and quantify the predictive performance of the models.ResultsThe machine learning algorithm with the best predictive performance is RF.The AUCs of the clinicoradiological model in training and validation cohorts were 0.793 and 0.701,respectively.The radiomics signature of arterial phase(AP)images alone achieved satisfying predictive efficacy for MVI,with AUCs of 0.671 and 0.643 in training and validation cohort,respectively.The combination of clinicoradiological factors and fusion radiomics signature of AP and VP images achieved AUCs of 0.824 and 0.801 in training and validation cohorts,0.812 and 0.805 in prospective validation and external validation cohorts,respectively.The hybrid model provided the best prediction results.The results of the Delong test revealed that there were statistically significant differences among the clinicoradiological-radiomics hybrid model,clinicoradiological model,and radiomics model(p<0.05).ConclusionThe combination of clinicoradiological factors and fusion radiomics signature of AP and VP images based on Gd-EOB-DTPA-enhanced MRI can effectively predict MVI. |
PDF Full Text Request