The Predictive Value Of Conventional MRI Combined With SWI Nomogram Model For Benign And Malignant Lesions Of Parotid Gland

Purpose: Explore the predictive value of the Nomogram model of conventional magnetic resonance imaging combined with Susceptibility weighted imaging(SWI)for benign and malignant lesions of the parotid gland.The clinical diagnosis efficiency of the differentiation of benign and malignant lesions of the parotid gland provides more reliable basis for the preoperative diagnosis of parotid gland lesions.Methods: Retrospectively analyzed the clinical and imaging data of 51 patients with parotid gland lesions in our hospital.All the cases were confirmed by postoperative pathology,including 28 benign lesions and 23 malignant lesions;MR examination and scanning were performed before treatment,including MR Routine sequence and SWI sequence to obtain the corresponding MR signs,including MR conventional signs: diameter of parotid gland lesions,the shape of the lesion,whether it involves the deep lobe,the edge of the lesion,whether there are swollen lymph nodes in the neck;SWAN related signs :Intratumoral Susceptibility Signal intensity(ITSS),distribution of veins in parotid gland lesions.Perform a single factor analysis on all MR signs of benign and malignant lesions of the parotid gland to obtain the characteristics of each sign distribution,perform logistic regression analysis based on the optimal subset criterion,and perform k-fold cross-validation(k = 10)to finally obtain the corresponding parameter values,Including: odds ratio(OR),95% CI and corresponding P value.Use R software to establish a nomogram prediction model,use discrimination and calibration to evaluate the performance of the prediction model,and use decision curve analysis,net reclassification index(NRI)and comprehensive discriminant index(IDI)to evaluate ITSS to predict the nomogram the contribution of the model.Results:(1)The results of univariate analysis of conventional MR signs suggested that there were statistical differences in lesion morphology,involvement of the deep parotid gland,uneven edges,and enlarged cervical lymph nodes(P <0.005).(2)Logistic regression analysis based on the optimal subset criterion showed that deep leaves(OR = 7.429,P = 0.042),lesion margins(OR = 5.572,P = 0.044),and ITSS(OR = 3.407,P = 0.012)is an independent risk factor for predicting malignant lesions of the parotid gland.(3)Based on Logistic regression,a Nomogram prediction model for predicting malignant lesions of the parotid gland was established.The consistency index of the model was C-index = 0.898(95% CI: 0.856-0.940),which was of moderate accuracy.The ROC curve showed that the AUC of the model was 0.898,which was sensitive The degree and specificity are 82.60% and 82.10% respectively;the calibration curve observes that the model predicted value and the actual observed value both fall near the 45°diagonal line,and the mean absolute error(Mean absolute error,MAE)is 0.049,indicating that the model Better clinical calibration.When quantifying the calibration degree,the Brier score obtained in this study was 0.131,which also reflected the better calibration ability of this model.Decision curve analysis,net reclassification index and comprehensive discriminant improvement index suggest that the addition of ITSS improves the predictive ability of the model.Conclusion: The Nomogram model of conventional magnetic resonance imaging combined with magnetic sensitivity weighted imaging(SWI)has a high accuracy in the prediction of benign and malignant lesions of the parotid gland,and is suitable for clinical practice.Among them,the magnetic sensitivity signal intensity(ITSS)can improve the predictive ability of the Nomogram model of benign and malignant lesions of the parotid gland.