Predicting The Recurrence Of Acute Pancreatitis Based On Clinical Features And Radiomics Model Of MR

Part I Predicting the recurrence of acute pancreatitis based on clinical and MR imaging featuresObjective:Based on the clinical and MR imaging characteristics of patients in non recurrent group and recurrent group of acute pancreatitis(AP),a combined risk model of clinical and imaging characteristics of recurrent acute pancreatitis(RAP)was established to provide reference basis for the prevention and diagnosis of RAP.Methods:According to the inclusion and exclusion criteria,218 patients with acute pancreatitis in our hospital from January 2015 to December 2017 were retrospectively collected,including 132 cases of non recurrent group(non-RAP group)and 86 cases of recurrent group(RAP group).The clinical characteristics and MR imaging features data of the two groups were recorded.Firstly,the data were analyzed by univariate analysis,and the characteristic with statistical differences between the two groups were further included in multivariate analysis.The independent risk factors of AP recurrence were analyzed,and a combined risk model of clinical and imaging characteristics of RAP was established by logistic regression.Result:There were significant differences in the ratio of age(P=0.001),gender(P<0.001)and incidence of hyperlipidemia(P<0.001)between the RAP group and non-RAP group.There was no significant difference in the ratio of drinking,biochemical indexes of pancreatitis between the two groups(all P>0.05).The severity of pancreatitis(P<0.001),magnetic resonance severity index(MRSI)(P=0.001),complication classification(P=0.012),type of pancreaticobiliary junction(P=0.003)and angle of pancreaticobiliary junction(P<0.001)were statistically different between the two groups.There was no significant difference in the incidence of bile duct stones(P>0.05).Logistic regression analysis showed that male patients,V-type and B-P-type junction,greater angle of pancreaticobiliary,hyperlipidemia and first-episode of AP as mild or moderate severe were risk factors for RAP,and younger age was protective factor for RAP.The comprehensive prediction accuracy of the combined risk model of clinical and imaging characteristics was 81.2%.Conclusion:Clinical features combined with MR imaging features can moderately predict the recurrence of acute pancreatitis and provide a more comprehensive basis for the individualized prevention,diagnosis and treatment of RAP.Part Ⅱ Predicting the recurrence of acute pancreatitis based on radiomics model of gadolinium contrast enhanced MRObjective:To investigate the value of radiomics model based on gadolinium contrast enhanced magnetic resonance imaging(CE-MRI)in predicting the recurrence of acute pancreatitis(AP).Methods:According to the inclusion and exclusion criteria,201 patients with acute pancreatitis in our hospital from January 2017 to December 2020 were retrospectively collected,including 102 cases of non recurrent group(non-RAP group)and 99 cases of recurrent group(RAP group).They were randomly divided into training set and verification set in the ratio of 7:3.The quantitative features of radiomics were extracted from the late MR contrast-enhanced arterial images,and the features with consistency test coefficient<0.75 were screened out.Then the feature selection and dimensionality reduction were carried out by univariate analysis and least absolute shrinkage and selection operator(LASSO)regression.Finally,the best subset of radiomics features retained after dimensionality reduction was used to construct the radiomics model by logistic regression analysis.The predictive ability of the model was evaluated by the receiver operating characteristic(ROC).Result:Among the patients in the training set,five best radiomics features were determined through feature selection and dimensionality reduction(3 gray level co-occurrence matrix features including cluster shadow and correlation,and 2 shape features including surface area density and voxel unit).The radiomics model has good prediction performance in the training set and verification set(AUC is 0.915 and 0.917,respectively).Conclusion:The radiomics model based on gadolinium contrast enhanced MR can moderately predict the recurrence of acute pancreatitis.It can become a new imaging marker for detection of RAP,and provide an objective basis for the optimization of prevention and treatment plan and prognosis evaluation of patients.