Multiparameter MRI Radiomics Model Predicts Preoperative Peritoneal Carcinomatosis In Ovarian Cancer

Objective: To explore the predictive value of radiomic features based on multi-parameter magnetic resonance imaging(multiparameter magnetic resonance imaging,MP-MRI)for peritoneal metastasis(peritoneal carcinomatosis,PC)of ovarian cancer(ovarian cancer,OC).Method: A retrospective collection of 86 patients with epithelial ovarian cancer who underwent MRI examinations in our hospital from June 2015 to May 2020 and were pathologically confirmed were included in this study,aged 32-82 years.All patients underwent conventional T2 WI,DWI,and DCE-MRI scans,and then underwent double-attachment hysterectomy plus omentectomy.Use ITK-SNAP software on the T2-weighted lipid pressure imaging(FS-T2WI),diffusion weighted imaging(DWI)and dynamic enhanced magnetic resonance imaging(DCE-MRI)images of each patient’s preoperative arterial late stage,step by step along the edge of the lesion The region of interest(region of interest,ROI)is drawn layer by layer,and finally fused into a three-dimensional volume block.Then,Py Radiomics is used for quantitative imaging feature extraction,and 1037 quantitative features are obtained.The intra-group correlation coefficient(intraclass correlation coefficient,ICC)is used to evaluate the feature repeatability,and the minimum redundancy is adopted.The maximum correlation(Max-Relevance and Min-Redundancy,m RMR)and minimum absolute shrinkage selection operator(LASSO)methods were used for feature screening,and 508,557 and 508 radiomic features were retained,respectively.The penalty parameter(λ)is adjusted through 10 times of cross-validation,and the optimal λ is found,so that the final value of λproduces the largest area under curve(area under curve,AUC),and the non-zero coefficient features related to the PC state are selected.Then multivariate Logistic regression was used to further screen the features,and finally a single model based on the three MRI sequences and a combined sequence model of the three were established to evaluate the ability of radiomics to distinguish PC.In addition,another model based on radiomic characteristics and clinicopathological risk factors was developed,and the receiver operating characteristic curve(ROC)was used to compare the diagnostic performance of different models.Finally,the multi-factor Logistic regression method was used to construct.Results: There were 39 patients with peritoneal metastasis confirmed by surgery and pathology,and 47 patients without peritoneal metastasis(age 33-82 years old,median age 54 years old).The radiomic model showed that the radiomic model from the MP-MRI combined sequence had a higher AUC than the model from FS-T2 WI,DWI and DCE-MRI alone,with AUC of 0.762(0.662-0.861)and 0.830(0.745-0.914),0.807(0.717-0.898),0.846(0.765-0.927),the differences were statistically significant.Preoperative CA125 level,DWI＿HLH＿glszm＿Size Zone Non Uniformity Normalzed,T1C＿glszm＿Low Gray Level Zone Emphasis,T1C＿LHL＿ngtdm＿Contrast are independent predictors,combined with radiomic characteristics and clinical pathological risk factors to construct a radiomic nomogram and clinical model0.840-0.,0.858(0.779-0.938),0.846(0.765-0.927),the AUC of the radiomics nomogram is higher than the clinical model and the omics model.Decision curve(DCA)analysis indicates the net benefit value under different risk thresholds.When the risk threshold is 37% to 85% in the training group,the net benefit value obtained by using radiomics to assess ovarian cancer peritoneal metastasis is greater than the total value.The net income value of intervention and non-intervention.The H-L test is not significant(p>0.05),indicating that it has good stability and does not deviate from the perfect fit.Conclusion: The arterial advanced MP-MRI radiomic analysis based on the primary tumor provides valuable information for predicting PC in ovarian cancer.The radiomic nomogram based on the MP-MRI combined sequence has a good predictive ability for ovarian cancer PC.This non-invasive and reliable tool can be used to identify peritoneal metastases in patients with ovarian cancer before surgery.