Preliminary Diagnosis Value Of Bladder Pathological Urothelial Carcinoma With Different Pathological Grades Based On CT Image Texture Analysis

Objective Based on CT image texture analysis,the value of initial diagnosis of bladder urothelial carcinoma patients with different pathological grades provides reference for clinical patients.Methods A retrospective analysis of 86 patients who underwent CT scans in our hospital from February 2017 to February 2019 due to clinically suspected bladder disease.The patients were 32 to 89 years old,74 males and 12 females.The surgical pathology of the patients admitted to the hospital was bladder cancer.According to different pathological grades of patients,they were divided into HGUC group(44 cases)and LGUC group(42 cases).All patients underwent pelvic CT scan and enhanced scan.The patient’s scan ranged from the level of the humeral condyle to the horizontal position of the lower edge of the pubic symphysis.The texture analysis of CT images was performed using the AK(Artificial Intelligent Kit)emission group analysis software developed by GE.The number of lesions,the generation of parameters and the image of the patient’s imaging were recorded.The differences between the texture images of the two groups were compared,including the root mean square,skewness,percentage of the area,cluster darkness,and large area and high gray level enhancement of the plain image.Enhance image kurtosis,entropy,area percentage,cluster darkness;draw the receiver operating characteristic(ROC)curve according to the patient image texture difference,and determine the optimal diagnostic boundary value of the texture parameter to evaluate the diagnostic effect.Results A total of 106 lesions were removed from 86 patients,including 52 patients with LGUC and 54 patients with HGUC.The length of the lesion ranged from 1.04 cmto 8.20 cm,with an average of(2.91 ± 0.64)cm.The lesions included pompon,cauliflower and cleavage..In this study,only 92 feature-based parameters are selected,including 60 first-order texture features,including 16 gray-scale shape matrices,16gray-scale length matrices,28 medium-matrix-level co-occurrence matrices,and 13 shape features.,19 statistic features.Two experienced imaging physicians independently analyzed the images and selected the ROI of the lesion’s region of interest,and calculated the parameters of the mass texture by post-processing software.Among them,the ICC(intra-group correlation coefficient)values of 92 texture parameters were scanned.The range is from 0.816 to 0.997,and the ICC(intra-group correlation coefficient)of 92 image texture parameters is enhanced from 0.855 to 0.998,and the consistency is good.The root mean square,skewness,regional percentage,and cluster darkness of HGUC patients were lower than those of LGUC group,and the large area and high gray scale enhancement was higher than that of LGUC group.The HGUC group enhanced image kurtosis,skewness,area percentage,and cluster darkness.The degree was lower than that in the LGUC group(P<0.05);the flatness index was 0.656,the best cut-off value was 0.188;the LGUC was diagnosed when the skewness was higher than 188,and the skewness was lower than 188.HGUC was diagnosed;the accuracy of HGUC for plain scan image was 83.01%,specificity was 73.10%,sensitivity was 92.60%,negative predictive value was 90.50%,and positive predictive value was 78.12%.Conclusion The texture analysis based on CT image is more accurate for the pathological level prediction of bladder cancer patients.The best diagnostic efficiency is the skewness of the plain scan image.The accuracy of the detection is as high as83.02%,which has certain clinical application value.