Abstract: Objective To explore the independent risk factors of early enteral nutrition aspiration in patients with severe acute pancreatitis (SAP) based on deep learning,and to establish a prediction model to predict the risk of early enteral nutrition aspiration. Methods The clinical data of 296 patients with severe acute pancreatitis in Tongji Hospital Affiliated to Tongji Medical College of Huazhong University of science and technology from January 2012 to December 2019 were retrospectively analyzed,including 268 patients without aspiration and 28 patients with aspiration.The clinical data of gender,age,body mass index,APACHE II score,consciousness,nutritional risk,nasogastric tube length,neutrophil-lymphocyte ratio (NLR) and platelet-lymphocyte ratio (PLR) were compared between the two groups Support vector machine (SVM) and generalized linear regression (GLR) algorithm were used to establish five prediction models to obtain the importance of prediction variables.At the same time,the subject work curve and decision curve were drawn to test the predictive value of the model. Results APACHE-II score,consciousness,nutritional risk,length of nasogastric feeding tube and PLR were the related variables to predict early enteral nutrition aspiration.The areas under the curve of random forest,neural network,decision tree,support vector machine and generalized linear regression algorithm were 0.976,0.973, 0.961 ,0.932 and 0.921,respectively.Through comparison,the performance of random forest algorithm was the best. Conclusion The prediction model based on machine learning algorithm can accurately predict the possibility of early enteral nutrition aspiration in patients with severe acute pancreatitis,which is conducive to postoperative evaluation and clinical nursing decision-making.

Key words: severe acute pancreatitis, enteral nutrition, risk of aspiration, machine learning, risk factors