Prediction and evaluation of the severity of acute respiratory distress syndrome following severe acute pancreatitis using an artificial neural network algorithm model

Abstract

Background: To predict the risk and severity of acute respiratory distress syndrome (ARDS) following severe acute pancreatitis (SAP) by artificial neural networks (ANNs) model. Methods: ANNs model was constructed by clinical data of 217 SAP patients. The model was first trained on 152 randomly chosen patients, validated and tested on the 33 patients and 32 patients respectively. Statistical analysis was used to assess the value of it. Results:The training, validation, and test set were not significantly different for 13 variables. After training, ANNs retained excellent pattern recognition ability. When ANNs model was applied to the test set, it revealed a sensitivity of 87.5%, and an accuracy of 84.43%. Significant differences were found between ANNs model and logistic regression model. When ANNs model is used to identify ARDS, the area under ROC was 0.859 + 0.048. Meanwhile, pancreatic necrosis rate, lactic dehydrogenase and oxyhemoglobin saturation were the most important independent variables. Compared with the Berlin definition, the ANN model shows a good accuracy of 73.1% for total severity of ARDS. Conclusion:ANNs model is a valuable tool in dealing with risk prediction of ARDS following SAP. In addition, it can extract informative risk factors of ARDS via the ANNs model.

Artificial intelligence in pancreaticobiliary endoscopy

“…A systematic review of neural network‐based methods noted that these methods have a sensitivity of 0.89 and specificity of 0.96 six hours after admission when compared with APACHE II (cutoff score ≥ 8) with 0.80 and 0.85, respectively 8 . Several other studies using AI‐based methods reported similar results 9,11–15 . AI‐based methods could also identify specific complications of AP such as porto‐venous thrombosis with an accuracy of 83% 16 and intra‐abdominal infections with a receiver operating characteristic curve of 0.923 [0.883–0.952] 17 …”

“…8 Several other studies using AI-based methods reported similar results. 9,[11][12][13][14][15] AI-based methods could also identify specific complications of AP such as porto-venous thrombosis with an accuracy of 83% 16 and intra-abdominal infections with a receiver operating characteristic curve of 0.923 [0.883-0.952]. 17 One of the earliest applications of AI methods in pancreatobiliary endoscopy is from 1998 where Yeaton et al used decision trees on image variables generated from brush cytology obtained during endoscopic retrograde cholangiopancreatography (ERCP) to distinguish between chronic pancreatitis and pancreatic adenocarcinoma (PDAC) with a sensitivity of 91% and specificity of 87%.…”

Artificial intelligence for the management of pancreatic diseases

“…Three studies aimed to predict complications by using an ANN and compared it to logistic regression (LR) modeling. The results showed that the ANN significantly outperformed the LR modeling in predicting the occurrence of several complications during the course of the disease in all three studies 17–19 . Two studies reported ANNs that predict multi‐organ failure (MOF) in AP patients based on clinical and laboratory findings.…”

“…Several studies report ANNs that predict complications and mortality in patients with AP with high accuracy, ranging from 83.0% to 97.5% 17–23 . Three studies aimed to predict complications by using an ANN and compared it to logistic regression (LR) modeling.…”

<p>Prediction of Lymph Node Metastasis in Superficial Esophageal Cancer Using a Pattern Recognition Neural Network</p>

“…In several previous papers, the discriminatory power of logistic regression (LR) and ANN models was compared. [9][10][11] Both models performed equally well in most cases, whereas the more flexible ANN model generally outperforming LR model in the remaining cases. Among the published prediction models for LN metastasis, there are already three LR-based nomograms.…”