Clinical Application Of Predicting Lymph-vascular Space Invasion In Cervical Cancer Based On MRI Radiomics And Clinical Radiology Characteristic Nomograms

Objective: To explore the clinical application value of nomograms based on clinical,radiology and MRI radiomics features in lymph-vascular space invasion（LVSI）of cervical cancer.Methods: A total of 247 patients with cervical cancer were enrolled in this study.The clinicopathological and radiology data of all patients were collected.Univariate and multivariate regression analysis were used to determine the independent risk factors that can predict LVSI of cervical cancer in clinical and radiology features.The region of interest（ROI）was manually delineated using Infer Vision platform and the radiomics features of axial T2WI-FS（T2ax）,DWI and sagittal T2WI-FS（T2sag）were extracted respectively,the patients were randomly divided into training group（n=172）and verification group（n=75）by IPMs and the most valuable radiomics features are selected through dimensionality reduction,and then the radiomics models are constructed respectively.The diagnostic efficacy of each radiomics models in the training group and verification group on the state of LVSI was evaluated by the receiver operating characteristic（ROC）curve,and the best model was selected and the radiomics score（Radscore）was calculated.Then based on independent clinical and radiology predictors and Radscore,a clinical-radiology nomogram was constructed to predict the LVSI of cervical cancer.Finally,Hosmer-Lemeshow test,calibration curve,ROC curve and decision curve analysis（DCA）are used to verify the performance of the nomogram.Results: 1.Univariate difference analysis showed that red blood cell（RBC）,squamous cell carcinoma antigen,differentiation,status of lymph node metastasis in MRI images,maximum lesion diameter and apparent diffusion coefficient（ADC）were correlated with LVSI status of cervical cancer（P<0.05）.However,multivariate Logistic regression analysis showed that only RBC,differentiation and ADC value were independent risk factors for LVSI in patients with cervical cancer（P<0.05）.2.The ROC curve showed that the area under the curve（AUC）,sensitivity and specificity of the radiomics model based on combined sequence（T2ax+T2sag+DWI）Radscore in the training group were 0.806,80.0% and 67.0%,respectively,and the corresponding values in the verification group were 0.728,71.4% and 66.0%,respectively.3.Clinical-radiology nomogram based on clinical and radiology independent risk factors and Radscore.The calibration curve shows that the predicted LVSI probability is in good agreement with the actual LVSI.The nomogram has the best prediction performance in the ROC curve.Its AUC,sensitivity and specificity in the training group were 0.895（95%CI:0.839-0.936）,82.7% and 84.5%,respectively,and the corresponding values in the verification group were 0.834（95%CI:0.730-0.910）,60.7%and 91.5%,respectively.The DCA curve shows that the net benefit of the clinicalradiomics nomogram is greater than 0 in the threshold probability range of 0-0.95.Conclusion: RBC,differentiation,ADC value and Radscore can be used as independent influencing factors of LVSI in cervical cancer.The clinical-radiomics nomogram constructed based on the above independent influencing factors has good predictive efficiency and clinical application value for LVSI status in cervical cancer patients.