The Value Of Texture Analysis Based On Plain And Contrast-enhanced MRI Images In Differential Diagnosis Of Hepatic Focal Lesions

Background:Magnetic resonance imaging is widely used in the differential diagnosis of focal liver lesions,but it is difficult to identify hepatic tumors with overlapping images by conventional magnetic resonance imaging.Texture analysis can quantitatively analyze the gray intensity and spatial position changes in the lesions Indirectly reflecting the histopathological changes of the lesion,it can extract the texture information inside the lesion with high throughput and accurately quantitatively evaluate the heterogeneity of the tumor.Texture analysis has shown great value in identify diagnosis and prognostic evaluation of conventional hepatic tumors.Objective: To investigate the value of texture analysis based on plain and contrastenhanced MRI images in the differential diagnosis of benign and malignant focal hepatic lesions.Methods: A retrospective analysis was conducted on 87 patients with focal liver lesions found by MRI examination in our hospital from January 2019 to September 2020.All patients had complete imaging and clinical medical records.The patients were divided into benign and malignant groups,including 60 males and 27 females,aged from 27 to 80 years,with an average age of 56.2±10.9years.The conventional T2 WI fat-suppressed images and the three-stage contrast-enhanced images were imported into Omni-Kinetics software to extract the texture parameters.Two independent sample t test and MannWhitney U test were used for statistical analysis.Parameters with statistically significant differences were used to two factors binary logistic regression model and the ROC curve was used to evaluate its effectiveness in predicting the benign and malignant liver focal lesions.Results: 1.The minimum,maximum,meanvalue,kurtosis,energy,and GLCMenergy of the benign group were higher than that of the malignant group,and the skewness,entropy,GLCMentropy of the benign group was lower than that of the malignant group in T2 WI FS sequence(P<0.05).The maximum,standarddeviation,skewness,kurtosis,energy,and GLCMenergy of the benign group were higher than that of the malignant group,and the entropy,uniformity,GLCMentropy of the benign group was lower than that of the malignant group in arterial phase(P<0.05).The maximum,meanvalue,standard deviation,energy,and GLCMenergy of the benign group were higher than that of the malignant group,and the correlation,GLCMentropy of the benign group was lower than that of the malignant group in portal venous phase(P<0.05).The minimum,maximum,meanvalue,standarddeviation,and GLCMenergy of the benign group were higher than that of the malignant group,and the correlation,kurtosis,GLCMentropy of the benign group was lower than that of the malignant group in the delayed phase(P<0.05).2.Logistic regression analysis showed that the meanvalue,entropy in T2 WI FS and the standarddeviation,entropy and GLCMenergy in the arterial phase and the standard deviation,GLCMenergy in portal venous phase and the meanvalue,GLCMentropy,correlation in delayed phase were independent prediction factors in the differentiation of benign and malignant liver lesions,The differential diagnostic efficiency of GLCMenergy in arterial phase was the highest,with AUC of 0.816,sensitivity and specificity was 80% and 88.89%,respectively.The binary logistic regression model in the delayed phase was the most significant diagnostic parameter for distinguishing benign and malignant hepatic tumors,with an AUC of 0.875,with sensitivity and specificity of 80.00% and 88.89%,respectively.3.The GLCMenergy of T2 WI FS and enhancement sequence,and the contrast of delayed phase was significantly lower in the hepatocellular carcinoma group than the liver metastasis group(P<0.05),while the uniformity of the portal venous phase and the correlation of T2 WI FS sequence and delayed phase were higher than liver metastasis group(P<0.05);Logistic regression analysis showed that the correlation of T2 WI FS sequence,the GLCMenergy and uniformity of the portal venous phase,and the contrast of the delayed phase were the Independent prediction factors in the differentiation of hepatocellular carcinoma and liver metastasis.The differential diagnostic efficiency of GLCMenergy in portal venous phase is higher,with AUC of 0.733,the sensitivity of 71.87%,and specificity of 73.08%;The area under the curve of the combination of the two parameters in portal venous phase was 0.764,significantly higher than any other parameters,with a sensitivity of 90.62%,specificity of 53.85%.Conclusion: The texture analysis based on MRI images is helpful for the differentiation of liver focal lesions.The texture analysis is expected to be an effective means for the qualitative diagnosis of focal liver lesions.