Clinical Application Of Contrast-enhanced Ultrasound Liver Imaging Reporting And Data System

Part 1 Contrast-Enhanced Ultrasound Liver Imaging Reporting and Data System in Diagnosing Hepatocellular Carcinoma: Diagnostic Performance and Interobserver AgreementBackground: The American College of Radiology has recently released a contrast-enhanced ultrasound(CEUS)liver imaging reporting and data system(CEUS-LI-RADS)for classifying lesions at risk for hepatocellular carcinoma(HCC).We attempted to determine the diagnostic performance and inter-reader agreement of CEUS-LI-RADS in diagnosing HCC in high-risk patients.Method: From May 2017 to January 2020,this retrospective study consecutively included 144 high risk HCC patients with 144 focal liver lesions(FLL)from the SecondAffiliated Hospital of Zhejiang University and the Second Affiliated Hospital of Harbin Medical University.All FLLs underwent preoperative CEUS examination and were confirmed by biopsy/postoperative pathological specimens.Four blinded readers from two tertiary centers assigned CEUS-LI-RADS categories [LR-5 for predicting HCC: arterial phase enhancement(APHE)and late and mild washout] and compared with definite HCC diagnosis(APHE with portal venous wash-out regardless of onset time or intensity)per 2017 China Liver Cancer guidelines(CLCG).Finally,the pathological results were used as the gold standard.The diagnostic performance of CEUS LR-5 and 2017 CLCG were calculated by the sensitivity,specificity,positive predictive value(PPV)and negative predictive value(NPV).Chi-square test was used for pairwise comparison.In addition,the intra-class correlation coefficient(ICC)was used to evaluate the degree of consistency between multi-readers.Result: A total of 104 HCC patients and 44 non-HCC patients were included in this study.As the main diagnostic criteria for CEUS LR-5,APHE plus late and mild washout were more common in HCC patients,with positive rates of 90.4%-94.2% and 51.0%-56.7%,respectively,which were much higher than those of non-HCC patients(both P<0.05).According to the interpretation of four sonographers,the prevalence of HCC in LR3,LR4,LR5,and LR-M was 17.3%,55.8%,94.2% and 57.9%,respectively.If CEUS LR-5 is used as the criteria for diagnosing HCC,the sensitivity,specificity and PPV of the four sonographers in diagnosing HCC would reach 46.2%-53.9%,90.9%-95.5%,93.0%-96.0%,respectively.If 2017 CLCG was used as the criteria for diagnosing HCC,the sensitivity,specificity and PPV of the four sonographers in diagnosing HCC would be 90.4%-93.9%,38.6%-43.2% and 77.7%-79.2%.Compared with 2017 CLCG,thespecificity and PPV of each physician’s CEUS LR-5 diagnosis were significantly increased(all P<0.05),but the sensitivity was significantly reduced(all P<0.05).For single imaging features,inexperienced radiologists had moderate agreement on late and mild washout,early washout,marked washout and mosaic signs,while agreement with rim enhancement and APHE were more consistent.On the other hand,experienced radiologists had moderate agreement on late and mild washout and marked washout,but the differences on APHE,rim enhancement,early washout,and mosaic signs were small.All readers had poor consitentcy for interpretating nodule-in-nodule sign,and the ICC value was less than 0.20.The inter-reader agreement of the above single imaging feature were not affected by experience level(all P>0.05).Concerning the final CEUS LI-RADS classification,the consistency between groups of inexperienced(ICC=0.56)or experienced readers(ICC=0.59)was relatively moderate.Conclusion: Among high-risk HCC patients,CEUC LR-5 has high specificity and PPV value in the diagnosis of HCC,which would minimize the possiability of misdiagnosing other non-HCC liver nodules as HCC.However,as an ancillary feature favor for HCC,nodule-in-nodule pattern obtained low diagnostic ability to distinguish HCC from non-HCC lesions.Furthermore,sonographers achived high inter-reader variabilities when interpretating nodule-in-nodule pattern.Only fair to moderate inter-reader agreement were achieved for CEUS LI-RADS final classification and washout appearance(late and mild washout and marked washout).Part 2 Influence of clinicopathological factors on HCC CEUS LI-RADS Background:Exploring the clinical or pathological factors that may affect the CEUS features of HCC lesions indicates clinical significance for further understanding the final classification of HCC lesions by CEUS LI-RADS.Therefore,this study focuses on exploring the correlation between clinicopathological information and CEUS features or final category.Method: From May 2017 to January 2020,this retrospective study included 88 HCC patients from the Second Affiliated Hospital of Zhejiang University and the Second Affiliated Hospital of Harbin Medical University.All patients were confirmed by postoperative pathological specimens.Clinicopathological information includes gender,age,tumor size,tumor cell differentiation degree,liver function indicators,coagulation indicators and abnormal tumor serum marker levels.Independent t-test or chi-square test were used to evaluate the distribution of clinicopathological information in CEUS features or final category.Result: In terms of single CEUS LI-RADS imaging features,patient’s gender,age,tumor size,tumor cell differentiation,liver function indicators,coagulation indicators,and increased tumor marker levels cannot effectively distinguish APHE from non-APHE,late and mild WO from non- late and mild WO,rim enhancement from non-rim enhancement,marked WO from non- marked WO,and nodule-in nodule from nodule-in nodule pattern(all P>0.05).Early WO(n=35)was found in 39.8% of all cases.Therewere no significant differences in the distribution of patient gender,age,liver function indicators,coagulation indicators,and increased tumor marker levels(all P>0.05)between early WO and non early WO group.However,tumor size>30mm was more common in early WO lesions(80.0% vs.47.2%,P=0.003).In addition,poorly differentiated tumor was also more frequently detected in early WO group,accounting for 51.4%,which was significantly higher than that in the non-early WO group(26.4%,P=0.024).Tumor size>30mm(P=0.001),poorly differentiated tumor(P=0.002)and albumin <35mg/L(P=0.016)mostly occurred in HCC lesions with mosaic pattern,and the differences were statistically different.Concerning final CEUS LI-RADS classification,poorly differentiated tumor were more common in LR-M HCC lesions(51.4%),and less common in LR-3/4(18.2%)and LR-5 groups(27.5%)(P=0.038).Conclusion: In terms of single CEUS LI-RADS features,several clinicopathological factors,such as the degree of HCC cell differentiation and tumor size,have limited influence on the imaging features of CEUS LR-5 category(APHE+late mild clearance).However,tumor cell differentiation and tumor size have significant impact on early washout and mosaic patterns.Compared with other CEUS LI-RADS classifications,LR-M HCC contains more poorly differentiated HCCs.Part 3 A Nomogram Combining Clinical Features and Contrast Enhanced Ultrasound LI-RADS for Preoperative Prediction of Microvascular Invasion in Hepatocellular Carcinoma PatientsBackground: Microvascular invasion(MVI)is one of the most important prognostic factor for predicting survival status of HCC patients.However,up to now,there was no relevant study focusing on the role of preoperative features on CEUS LI-RADS in predicting MVI.The aim of this study is to establish a predictive model incorporating clinical features and contrast enhanced ultrasound liver imaging and reporting and data system(CEUS LI-RADS)for estimation of MVI in HCC patients.Methods: In the retrospective study,127 HCC patients from the Second affiliated Hospital of Zhejiang University School of Medicine and the Second affiliated Hospital of Harbin Medical University were allocated as training cohort(n=98)and test cohort(n=29)based on cutoff time-point,June 2020.Multivariate regression analysis was used to select independent indicators for developing predictive nomogram models.We also performed area under receiver operating characteristics(AUC)curve to indicate diagnostic performance of different predictive models.The corresponding sensitivity and specificity of different models at the cutoff nomogram value were compared.Results: In the training cohort,clinical information(larger tumor size,higher AFP level)and CEUS LR-M are significantly correlated with the presence of MVI(all P<0.05).Byincorporating clinical information and CEUS LR-M,the predictive model(LR-M+Clin)achieved desirable diagnostic performance(AUCC=0.80 and 0.84)in both cohorts at a cutoff value of nomogram score 89.In the test group,the clinical feature prediction model alone(Clin)has an AUC of 0.66 for MVI-positive patients with sensitivity and specificity of 46.7%,and 85.7%,respectively.In comparison,the sensitivity of LR-M+ Clin for predicting MVI in HCC patients,was higher in comparison with that of Clin model alone(86.7% vs.46.7%,P=0.027)without losing specificity(78.6% vs.85.7%,P=0.06)in the test cohort.In addition,LR-M+Clin exhibited similar AUC and specificity,but a significantly higher sensitivity(86.7%)than those of LR-M alone and LR-5(No)+Clin(both sensitivities=73.3%,both P=0.048).Conclusion: The present predictive model incorporating CEUS LR-M and clinical features were able to predict MVI status of HCC with desirable diagnostic performance and might serve as a reliable preoperative tool for guiding further treatment algorithm.