Using intelligence techniques to predict postoperative morbidity of endovascular aneurysm repair

Abstract

Endovascular aneurysm repair (EVAR) is an advanced minimally invasive surgical technology that is helpful for reducing patients' recovery time, postoperative mortality and morbidity. This study proposes an ensemble model to predict postoperative morbidity after EVAR. The ensemble model was developed using a training set of consecutive patients who underwent EVAR between 2000 and 2009. All data required for prediction modeling, including patient demographics, preoperative, co-morbidities, and complication as outcome variables, was collected prospectively and entered into a clinical database. A discretization approach was used to categorize numerical values into informative feature space. The research outcomes consisted of an ensemble model to predict postoperative morbidity, the occurrence of postoperative complications prospectively recorded, and the causal-effect decision rules. The probabilities of complication calculated by the model were compared to the actual occurrence of complications and a receiver operating characteristic (ROC) curve was used to evaluate the accuracy of postoperative morbidity prediction. In this series, the ensemble of Bayesian network (BN), artificial neural network (ANN) and support vector machine (SVM) models offered satisfactory performance in predicting postoperative morbidity after EVAR. © 2011 Springer Berlin Heidelberg.