The Value Of Nomogram Based On MRI Radiomics And Clinical Imaging Characteristics In Predicting Pancreatic Fistula After Pancreaticoduodenectomy

Objective:To explore the value of nomogram based on MRI radiomics and clinical imaging characteristics in predicting pancreatic fistula（PF）after pancreaticoduodenectomy（PD）.Methods:The clinical and MRI data of 134 patients who underwent PD treatment and abdominal MRI examination in the First Hospital of Lanzhou University from January 2017 to August 2021 were retrospectively collected.Univariate analysis and multivariate logistic regression analysis were used to determine the independent risk factors for predicting POPF in clinical imaging features,which were used to construct clinical models.On axial T₁WI and DWI images,the region of interest（ROI）was delineated layer by layer along the pancreatic parenchymal region and the corresponding radiomics features were extracted.The patients were randomly divided into training set（n=93）and validation set（n=41）.Spearman correlation analysis（SPM）,Mann-Whitney U test and minimum absolute shrinkage and selection algorithm（LASSO）were used to reduce the dimension and screen the features.Two machine learning models were constructed according to the selected features,and the optimal machine learning model was selected and its radiomics signature was calculated.Then,a clinical-radiomics model was established based on clinical imaging independent risk factors and signature,and a nomogram was drawn.The area under the receiver operating characteristic（ROC）curve（AUC）was used to evaluate the predictive performance of the nomogram.The calibration curve was used to evaluate the consistency between the predicted probability and the actual probability.The decision curve analysis（DCA）was used to evaluate its clinical application value.Results:The results of univariate and multivariate analysis showed that preoperative serum CA19-9≥325U/ml,main pancreatic duct diameter≤3mm,T₁WI signal intensity ratio（T₁WI SIR）≥1.32 were independent risk factors for POPF in PD patients（P<0.05）.The AUC,sensitivity and specificity of the clinical model constructed by 0.761,75.2%and 67.2%in the training set,and 0.631,50.0%and75.9%in the validation set.The AUC,sensitivity and specificity of the radiomics model based on T₁WI images were 0.794,93.8%and 59.0%in the training set,and0.809,83.3%and 69.0%in the validation set.The AUC,sensitivity and specificity of the clinical-radiomics model based on clinical imaging independent risk factors and signature were 0.861,81.2%and 80.3%in the training set,and 0.830,91.7%and62.1%in the validation set.The ROC curve showed that the nomogram had good predictive performance.The calibration curve showed that the predictive probability of the nomogram was highly consistent with the actual probability.The DCA curve showed that the nomogram had high clinical application value.Conclusion:The clinical-radiomics nomogram based on preoperative serum CA19-9 level,main pancreatic duct diameter,T₁WI SIR and signature has certain value in predicting the occurrence of POPF in PD patients.