Study On The Diagnostic Value Of Multiparameter MR Imaging Features In Parotid Pleomorphic Adenoma And Adenoma Lymphoma

Objective:To investigate the application of MR radiomics in the differentiation of parotid pleomorphic adenoma and adenomatous lymphoma.Materials and Methods:This study is a retrospective study.A total of 70 patients with benign parotid tumors were collected,including 38 patients with pleomorphic adenoma and 32 patients with adenomatous lymphoma.The imaging data and clinicopathological information of all patients were complete.(1)MRI images of the lesion(T2WI-FS,ADC)were manually sketched layer by layer to extract the area of interest(VOI),and1288 imaging radiomics features(first-order features,morphological features,texture features and advanced features)were extracted by means of imaging radiomics method.Dimension reduction was carried out for these features through feature Correlation analysis and linear Correlation Coefficient(Pearson Correlation Coefficient).Finally,the machine learning method of Logistic Regression was used to establish the imaging radiomics model based on T2WI-FS and ADC images by 10 fold cross validation.(2)Collect the clinical data of the patients(age,sex,smoking)and subjective image data(lesion size,number,location,position,morphology,boundary,with or without coated,T1 WI and T2WI-FS signal strength,and the mean ADC),single factor and multiple factor regression analysis were used to screen out the independent factors influencing the differential diagnosis of parotid pleomorphic adenoma and adenomatous lymphoma and to construct the clinical model.(3)Based on the previously constructed imaging omics model and clinical model,a fusion model of the two was constructed and a nomogram was drawn.The predictive value between models was assessed using the Receiver Operating Characteristic curve(ROC),while validation groups were used to validate the constructed models.Finally,the clinical practicality of clinical model,imaging radiomics model and combination model was observed by decision analysis curve.Results:The AUC of the imaging radiomics model based on T2WI-FS in the training group and the verification group were 0.915(95%CI:0.848-0.983)and0.891(95%CI:0.813-0.980),the accuracy was 0.857 and 0.857,the sensitivity was0.868 and 0.868,the specificity was 0.844 and 0.844,the positive predictive values were 0.868 and 0.868,the negative predictive values were 0.844 and 0.844,respectively.The AUC of the imaging radiomics model based on ADC in the training group and the verification group were 0.958(95%CI:0.916-1.000)and 0.907(95%CI:0.831-0.984),the accuracy was 0.871 and 0.857,the sensitivity was 0.868 and 0.868,the specificity was 0.875 and 0.844,the positive predictive values were0.892 and 0.868,the negative predictive values were 0.849 and 0.844,respectively.Statistical analysis showed that the mean value of ADC was an independent influencing factor for the differentiation of pleomorphic adenoma and adenolymphoma of the parotid gland,and a clinical model was established.Results showed that in the training and validation set AUC were 0.803(95% CI: 0.685 0.920),0.794(95% CI: 0.673 0.914),The accuracy was 0.771 and 0.771,the sensitivity was0.763 and 0.763,the specificity was 0.781 and 0.781,the positive predictive values were 0.806 and 0.806,the negative predictive values were 0.735 and 0.735,respectively.The AUC of the combined model training group and the validation group were 0.961(95%CI:0.918-1.000)and 0.944(95%CI:0.887-1.000),the accuracy was 0.914 and 0.900,the sensitivity was 0.895 and 0.895,the specificity was 0.937 and 0.906,the positive predictive values were 0.944 and 0.919,the negative predictive values were 0.882 and 0.879,respectively.The decision curve shows that the imaging radiomics model and the combination model have better clinical value than the clinical model.Conclusion:1、Imaging radiomics model constructed based on MRI T2WI-FS and ADC images has a high diagnostic efficacy in differentiating pleomorphic adenoma and adenolymphoma of the parotid gland.2、Normogram,constructed by radiomics model combined with clinical information,can be used for individualized prediction of pleomorphic adenoma and adenolymphoma of the parotid gland.Compared with the clinical model and radiomics model constructed by T2WI-FS,the prediction efficiency is significantly improved and has statistical significance.