The Value Of CT Finding And CT Texture Analysis In Predicting The Degree Of Pathological Differentiation Of Single Hepatocellular Carcinoma

Objective:To investigate the relationship between CT findings and pathological differentiation of hepatocellular carcinoma(HCC);The predictive value of texture analysis based on CT image in the degree of pathological differentiation of HCC;And whether quantitative CT texture analysis is better than CT findings in predicting the degree of pathological differentiation of HCC.Methods:(1)A total of 115 patients with single HCC treated in The People’s Hospital of Guangxi Zhuang Autonomous Region from October 2016 to September 2021 with complete clinical,pathological and imaging data were collected as the research objects.All patients underwent plain CT scanning and three-phase enhanced scanning(arterial phase,portal vein phase and balance phase)within 2 weeks before operation.According to the postoperative pathological results,they were divided into high differentiation and low differentiation groups.χ~2test was used to analyse the relationship between CT findings of HCC and its degree of pathological differentiation.(2)The preoperative original CT images of all patients were imported into Mazda4.6.After manually sketching the region of interest(ROI)layer by layer for HCC lesions,the software melts it into volume of interest(VOI)and extracts 94 texture features,which are processed with Mazda4.6.The feature filtering function of the software reduces the feature dimension by Fisher coefficient(Fisher),classification error probability combined average correlation coefficients(POE+ACC)and mutual information(MI).Finally,The B11 program in Mazda4.6 software calculates misclassification rate(MCR)of texture analysis based on CT image in the prediction of HCC pathological differentiation by four algorithms:raw data analysis(RDA),principal component analysis(PCA),linear discriminant analysis(LDA)and nonlinear discriminant analysis(NDA).(3)The results of CT findings with predictive value for the degree of HCC pathological differentiation and texture analysis combination(combination of feature filtering and MCR algorithm)with the best prediction effect on the degree of HCC pathological differentiation were counted,and these two data were compared by mcnemar’sχ~2test.Results:(1)Among 115 patients with HCC,the longest tumor diameter,pseudocapsule,tumor vessels and necrosis/cystic change on preoperative CT were correlated with the degree of pathological differentiation of HCC(P<0.05);In HCC with the longest diameter>3.0cm,the proportion of low differentiation was 65.93%,which was significantly higher than 29.17%of patients with the longest diameter≤3.0cm(P<0.05);In HCC with intact pseudocapsule,the proportion of low and middle grade patients was 45.95%,which was significantly lower than 80.49%of patients without or incomplete pseudocapsule(P<0.05);The proportion of poorly differentiated HCC with tumor vessels was 76.36%,which was significantly higher than 41.67%of HCC without tumor vessels(P<0.05);The proportion of moderately differentiated HCC with necrosis/cystic transformation(66.67%)was significantly higher than that of HCC without necrosis/cystic transformation(44.19%)(P<0.05).There was no significant difference between the two groups in the presence of liver cirrhosis,portal vein tumor thrombus and lymphadenopathy on CT(P>0.05).(2)In CT texture analysis,NDA is the best algorithm for predicting the degree of HCC pathological differentiation in all scanning phases.The MCR(14.78%～21.74%)predicted by enhanced CT texture analysis in this algorithm is lower than that of plain scan image(MCR is 22.61%～26.96%),and the MCR(14.78%)combined with MI+NDA texture analysis of arterial phase image is the lowest.(3)Among the predictive effects of CT imaging findings and CT texture analysis on the degree of pathological differentiation of HCC,the combination of MI+NDA texture analysis in arterial phase with the lowest MCR was better than the predictive effects of CT findings on the longest diameter of tumor,pseudocapsule,tumor vessels and necrosis/cystic change(all P<0.05).Conclusion:(1)The CT findings of HCC can reflect the malignant degree of the tumor to a certain extent.The longest diameter of the tumor>3.0cm,no or incomplete pseudocapsule,tumor blood vessels and necrosis/cystic transformation on CT are the risk factors of low and medium differentiated HCC.(2)Texture analysis based on CT images has certain value in predicting the degree of pathological differentiation of HCC,especially the combination of MI+NDA texture analysis of arterial phase images.(3)The combination of MI+NDA texture analysis in CT arterial phase with the lowest MCR is better than traditional CT finding in predicting the degree of pathological differentiation of HCC.