Outcome Analysis of Patients with Acute Pancreatitis by Using an Artificial Neural Network

Abstract

Rationale and Objectives. The authors performed this study to evaluate the ability of an artificial neural network (ANN) that uses radiologic and laboratory data to predict the outcome in patients with acute pancreatitis. Materials andMethods. An ANN was constructed with data from 92 patients with acute pancreatitis who underwent computed tomography (CT). Input nodes included clinical, laboratory, and CT data. The ANN was trained and tested by using a round-robin technique, and the performance of the ANN was compared with that of linear discriminant analysis and Ranson and Balthazar grading systems by using receiver operating characteristic analysis. The length of hospital stay was used as an outcome measure.Results. Hospital stay ranged from 0 to 45 days, with a mean of 8.4 days. The hospital stay was shorter than the mean for 62 patients and longer than the mean for 30. The 23 input features were reduced by using stepwise linear discriminant analysis, and an ANN was developed with the six most statistically significant parameters (blood pressure, extent of inflammation, fluid aspiration, serum creatinine level, serum calcium level, and the presence of concurrent severe illness). With these features, the ANN successfully predicted whether the patient would exceed the mean length of stay (A z ϭ 0.83 Ϯ 0.05). Although the A z performance of the ANN was statistically significantly better than that of the Ranson (A z ϭ 0.68 Ϯ 0.06, P Ͻ .02) and Balthazar (A z ϭ 0.62 Ϯ 0.06, P Ͻ .003) grades, it was not significantly better than that of linear discriminant analysis (A z ϭ 0.82 Ϯ 0.05, P ϭ .53).Conclusion. An ANN may be useful for predicting outcome in patients with acute pancreatitis. Key Words. Computers, diagnostic aid; computers, ANN; pancreatitis. © AUR, 2002An episode of acute pancreatitis may run a variable and unpredictable course (1,2). Prediction of the likely severity of an episode in an individual patient is difficult but has implications for the institution of therapy and for patient information. For this reason, several predictive systems have been devised. Some of these systems are based on clinical criteria (eg, the Ranson and the acute physiology and chronic health evaluation [APACHE] scoring systems [2-4]), whereas others, such as the Balthazar system (5,6), are based on radiologic features at computed tomography (CT) (Fig 1). To our knowledge, no system has been devised in which both clinical and radiologic indicators are combined to predict severity or outcome. In this study, this task was investigated by using two multivariate decision model approaches: an artificial neural network (ANN) and linear discriminant analysis.ANNs are computer models composed of parallel, nonlinear computational elements arranged in layers with a

Prediction of Severe Acute Pancreatitis at Admission to Hospital Using Artificial Neural Networks

“…In the field of AP, only a few studies using ANNs as predictive systems have been published. The results have been ambiguous, but promising results have also been presented [12,13,14,15,16], especially by Mofidi et al [14]. …”

Comparative Effectiveness of Total Population versus Disease-Specific Neural Network Models in Predicting Medical Costs

“…The four studies cited immediately above all compared neural network analyses with other analytic techniques. In the study predicting acute pancreatitis patient outcomes by Keogan et al, 14 neural network predictions were not significantly better than linear discriminant analysis predictions (C ϭ 0.83 and C ϭ 0.82, respectively). On the other hand, in the study of postanesthesia care unit length of stay by Kim et al, 15 a neural network predicted correctly in 81.4% of situations, while logistic regression analysis predicted correctly in 65.0%.…”

“…Neural networks have recently been used to predict, among other outcomes, acute pancreatitis patient outcomes, 14 length of stay in a postanesthesia care unit, 15 breast cancer survival, 16 and 5-year colon carcinoma survival. 13 One important issue is how neural network techniques compare in predictive power with statistical techniques, ie, linear and logistic regression.…”

Machine Learning Can Improve Prediction of Severity in Acute Pancreatitis Using Admission Values of APACHE II Score and C-Reactive Protein

“…Keogan et al [54] performed their study primarily to evaluate the ability of an artifi cial network that used radiological and laboratory data to predict outcome, focusing on CT as a predictive measure. CT is of value in diagnosis and prediction in pancreatitis [55] , but is not used routinely in the early stages in many centres.…”