Study On The Value Of T2WI Sequence Imaging In Predicting The Recurrence Of Acute Pancreatitis

Objective:To investigate the value of MR T2-weight imaging radiomics in predicting the recurrence of acute pancreatitis.Method:The study retrospectively collected 200 acute pancreatitis of our hospital from January 2016 to December 2020.There were 127 cases of initial AP and 73 cases of recurrent acute pancreatitis(recurrent acute pancreatitis,RAP).They were randomly divided into training group and verification group in the ratio of 7:3.139 cases in the training group(88 cases of AP,51 cases of RAP)and 61 cases in the verification group(39 cases of AP,22 cases of RAP).At the same time,the clinical characteristics of the two groups(age,gender,stones,hyperlipidemia,drinking history,complications and severity)were collected.Two radiologists with 5 years of experience in abdominal diagnosis used blind method to outline ROI around the pancreas(including pancreatic necrosis and avoiding common bile duct and blood vessels).Four common feature groups are extracted from the region of interest,namely gray level co occurrence matrix,gray level run length matrix,histogram and shape.Single factor analysis and lasso(least absolute shrinkage and selection operator,LASSO)algorithm were used for feature screening.Spearman correlation was used to analyze the correlation between the best radiomics features and clinical features.Then different algorithms were used to establish the radiomics model for predicting AP recurrence,and the clinical model and the combined model based on clinical characteristics and imaging characteristics were established by logistic regression.The prediction ability of the model is evaluated by area under the curve(AUC).Finally,nomogram was used to evaluate the recurrence risk of AP patients.Results:The AUCs of decision tree and logistic regression radiomics model in predicting AP recurrence in the training group were 0.873(95%CI:0.812-0.933)and 0.793(95%CI:0.712-0.874),respectively.The sensitivity was 0.863 and 0.647,and the specificity was 0.807 and 0.875,respectively.In the validation group,the AUCs were 0.702(95%CI:0.564-0.839)and 0.711(95%CI:0.570-0.852),respectively.The sensitivity was 0.545 and 0.455,respectively.The specificity was 0.821 and 0.949,respectively.The AUC of clinical model in predicting AP recurrence in the training group was 0.662(95%CI:0.5827-0.7419),the sensitivity was 0.745 and the specificity was 0.580.In the validation group,The AUC was 0.714(95%CI:0.6062-0.8216),the sensitivity was 0.864 and the specificity was 0.564.The AUC of the combined model in predicting AP recurrence in the training group was 0.818(95%CI:0.7415-0.894),the sensitivity was 0.725 and the specificity was 0.875.In the validation group,the AUC was 0.843(95%CI:0.7352-0.9515),the sensitivity was 0.864 and the specificity was 0.789.The combined model and the clinical model compare the ROC curve in the training group(P<0.01),and there was a trend of statistical difference in the validation group(P=0.059).Complication is correlated with range,total mean,and quartile(P<0.01),and is positively correlated,with r values of 0.606,0.606,0.606,respectively.There is a correlation between stones and the advantage of short-run low gray scale(P<0.01),and is positive correlation,with an r value of 0.575.Hyperlipidemia is correlated with the total mean and quartile(P<0.01),and is positively correlated,with r values of 0.501 and 0.575,respectively.Alcohol consumption is correlated with range,total average,and quartile(P<0.01),and is positively correlated,with r values of 0.699,0.822,0.699,respectively.Conclusions:Radiomics is effective in predicting the recurrence of acute pancreatitis,and can provide guidance for clinical treatment of acute pancreatitis.Hyperlipidemia can be used as an independent risk factor for predicting AP recurrence.The combined model based on imaging features and clinical features is better than the clinical model.