The Study Of 3.0 T Multi-parameter MRI Combined With Texture Analysis In Preoperative Prediction Of Cervical Cancer Subtypes And Ki-67 Expression

Explore the Value of IVIM-DWI Parameters and Texture Features in the Preoperative Evaluation of Different Subtypes of Cervical CancerObjective To investigate the value of IVIM-DWIand texture analysis（TA）in the differential diagnosis of different subtypes of cervical cancer.Methods A retrospective analysis of 60 cases with cervical carcinoma,including 38 cases of SCC,16 cases of AC,and 6 cases of SCCC.All patients underwent MRI examination before surgery,and underwent Muiti B 10 b DWI（IVIM-DWI）scanning.The whole lesion of MRI T2WI image was sketched manually,and calculate its texture features,including mean value,standard deviation,kurtosis,skewness,entropy and uniformity.The receiver operating characteristic curve（ROC）was drawn,and the area under the curve（AUC）was calculated.Results In the study of IVIM-DWI,the difference of ADC,D value between SCC and SCCC was statistically significant（P<0.001,AUC=0.867;P=0.001,AUC=0.920）,the difference of D,D^*value between SCC and AC was statistically significant（P=0.024,AUC=0.774;P=0.023,AUC=0.753）,and f value had statistical significance in differentiating AC from SCCC（P=0.034,AUC=0.850）.In the study of TA,the difference in entropy between AC and SCCC,kurtosis between SCC and AC,skewness between SCC and SCCC were statistically significant（P<0.001,AUC=0.875;P=0.027,AUC=0.910;P=0.020,AUC=0.766）.Conclusion IVIM-DWI or TA features is of limited value in preoperative evaluation of three histological subtypes of cervical carcinoma,providing objective imaging basis for personalized treatment of cervical cancer.Prediction of Ki-67 PI in Cervical Squamous Cell Carcinoma based on MRI T₂WI Texture Analysis: a Preliminary StudyObjective To explore the relationship between texture features of cervical squamous cell carcinoma（CSCC）based on MRI T₂WI and its Ki-67 proliferation index（PI）.Methods A retrospective study of 70 patients with CSCC confirmed by pathology,they were divided into low-expression group（Ki-67 < 50%）and high-expression group（Ki-67 ≥ 50%）according to Ki-67 PI.Region of interest（ROI）was manually sketched on the largest section of tumor on the axial T₂WI,and then imported ROIs to GE Analysis Kit（AK）software to obtain all texture features.Results There were 48 in training set（11 with low-expression and 37 with high-expression）,22 in testing set（5 with low-expression and 17 with high-expression）.A total of 397 texture features were obtained by AK software,of which 4 texture features were selected between the two groups by Spearman correlation and LASSO method,including Long run high grey-level emphasis（LRHGLE）,Cluster shade,GLCM entropy and Haralick correlation.The accuracy,specificity and sensitivity of the training and testing set calculated by Bayes method were 87.5%,72.7%,91.9% and 90.9%,80.0%,94.1%,respectively.The diagnostic efficiency（i.e.AUC）of predicting high and low Ki-67 PI using texture features of the training set and testing set was 0.910,0.899,respectively.Conclusion Texture analysis based on MRI T₂WI can noninvasively predict the high and low expression of Ki-67 in cervical squamous cell carcinoma before surgery,make up for the deficiency of pathological biopsy,help to evaluate its malignant degree and predict its recurrence.