Title: Development of a model for trauma outcome prediction: a real-data comparison of artificial neural networks, logistic regression and data mining techniques

Authors: C. Koukouvinos; C. Parpoula

Addresses: Department of Mathematics National Technical University of Athens, Zografou, Athens 15773, Greece ' Department of Mathematics National Technical University of Athens, Zografou, Athens 15773, Greece

Abstract: Data Mining (DM) plays an essential role in exploring and interpreting massive medical data sets. In this paper, data-mining techniques are performed for trauma data analysis. Three non-parametric classifiers (Multilayer Perceptron (MLP), Radial Basis Function Neural Networks (RBFN), Classification and Regression Trees (C&RT)) were applied for classification and prediction of data and were compared with Logistic Regression (LR) in terms of several performance measures. Furthermore, valuable clusters of records were identified through TwoStep algorithm, and used for accurate decomposition of the data into categories and for providing more specific information for trauma management. The purpose of this study was to design a predictive model to enable successful trauma outcome prediction. Some of the most commonly used predictive models that assess probability of survival for trauma victims are the Trauma and Injury Severity Score (TRISS) methodology and the Revised Trauma Score (RTS) which derive from specific input variables. The generalisation ability of these predictive models was tested and the derived results were discussed in terms of the sensitivity analysis performed on the best model. The development of the Neural Networks (NNs) approach revealed encouraging results as NNs, most of the times, outperformed the traditional approaches in terms of better models' predictability.

Keywords: data mining; ANNs; artificial neural networks; classification trees; clustering; logistic regression; medical data; modelling; trauma outcome prediction; trauma data analysis; trauma management.

DOI: 10.1504/IJBET.2012.049327

International Journal of Biomedical Engineering and Technology, 2012 Vol.10 No.1, pp.84 - 99

Published online: 12 Dec 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article