Research Of Multiparametric Magnetic Resonance And Radiomics Parameters In Predicting The Sensitivity Of Concurrent Chemoradiotherapy For Patients With Locally Advanced Cervical Cancer

Objective:To investigate the feasibility and value of intravoxel incoherent motion diffusion weighted imaging(IVIM-DWI)and radiomic parameters in predicting the sensitivity of concurrent chemoradiotherapy(CCRT)in patients with locally advanced cervical cancer(LACC),and to establish a nomogram model based on clinical and imaging parameters to provide early prognostic markers for the assessment of first-line treatment in patients with LACC.Methods:Data were retrospectively collected from 163 patients with pathologically confirmedⅠB～IVA cervical cancer in our hospital(Anhui Provincial Hospital Affiliated to Anhui Medical University),with the admission period spanning from 1 January 2015 to 1 January 2021.All enrolled patients underwent conventional MRI and IVIM-DWI scans within 2 weeks prior to CCRT,followed by standard concurrent chemoradiotherapy in our radiotherapy department,and conventional MRI was repeated 6 months after the end of treatment.While all enrolled patients were randomized 2:1 into training set(for model construction)and validation set(for efficacy testing),and then were further divided into sensitive group(complete remission)and resistant group(incomplete remission)based on tumor regression as shown by MRI 6months after CCRT.General clinical information was collected and collated from enrolled patients including age,maximum tumor diameter(imaging),squamous cell carcinoma antigen(SCC),International Federation of Gynecology and Obstetrics(FIGO)staging,lymph node status(imaging),histological staging and various inflammatory indicators such as neutrophils,monocytes,lymphocytes,platelets,hemoglobin,albumin,NLR,PLR,MLR,SIRI,PNI and AISI.Quantitative IVIM-DWI parameter values(ADC,D,D~*,f)and three-dimensional radiomic parameters(T2WI fat suppression sequence)were measured from the primary lesions before CCRT in all patients.The least absolute shrinkage and selection operator(LASSO)algorithm was used for initial screening of the radiomic parameters.The diagnostic value of all parameters was assessed by chi-square test,two independent samples t-test and Mann-Whitney U-test.A prediction model with nomogram scoring system was constructed based on multivariate logistic regression analysis.Calibration curves,decision curves and receiver operating characteristic(ROC)curves were plotted,and the p-value of Hosmer-Lemeshow test,area under the curve(AUC)and consistency index(C-index)were calculated to assess the diagnostic efficacy and stability of the nomogram model.Finally,the model was internally validated by 200 times ten-fold cross-validation and further external validation was performed by substituting the model into the validation set.Results:A total of 163 patients were included in this study,with 109 patients in the training set and 54 patients in the validation set.There were no statistical differences in general clinical data between the two sets,indicating that the grouping was random and reasonable.According to the review of MRI 6 months after the end of CCRT,there were 115 patients in the sensitive group and 48 patients in the resistant group.Among the clinical predictors only FIGO stage exhibited predictive value for the sensitivity of CCRT in patients with LACC,and all inflammatory indicators showed no significant predictive value.Among the IVIM-DWI parameters,the D and f values,in the training set,were significantly different between the sensitive and resistant groups,with lower D value and higher f value in the sensitive group.While f value showed an independent diagnostic value for the sensitivity of CCRT in LACC patients with an HR of 0.011,and its AUCs in the training and validation sets were 0.764 and 0.783,respectively.Among the radiomic parameters,original-glszm-Inverse Variance(IV),original-glszm-Size Zone Non Uniformity(SZN)and original-firstorder-Minimum(Minimum),in the training set,showed independent diagnostic significance in the multivariate logistic regression analysis with HRs of 0.322,1.014 and 0.984,respectively.Their AUCs in the training and validation sets were 0.706,0.869,0.850 and 0.724,0.885,0.828respectively.The parameters with statistical difference in univariate analysis were further incorporated into a multivariate logistic regression analysis and a nomogram was constructed based on the independent predictors.The AUC of the model in the training set was 0.987,with an average AUC of 0.985,a Brier score of 0.042 and a C-index of0.987 for 200 times ten-fold cross-validation.The model was further taken into the validation set with an AUC of 0.984,indicating high accuracy and stability of nomogram.Conclusions:The IVIM-DWI parameters and MRI-derived radiomic parameters of primary lesions in cervical cancer patients,to a certain extent,can reflect the microscopic information of water molecule diffusion,blood perfusion and intrinsic specificity in the tumor tissue,helping to predict the sensitivity of CCRT for patients with locally advanced cervical cancer.The combined application of IVIM-DWI,radiomic and clinical predictors can complement each other and further improve the predictive efficacy,providing an objective basis for predicting the efficacy and individualized treatment of patients with locally advanced cervical cancer.