Prediction of mortality in stroke patients using multilayer perceptron neural networks

We aim ed to predict mortality in stroke patients by using multilayer perceptron (MLP) neural networks. Materials and methods: A data set consisting of 584 stroke patients was analyzed using MLP neural networks. The effect of prognostic factors (age, hospitalization time, sex, hypertension, atrial fibrillation, embolism, stroke type, infection, diabetes mellitus, and ischemic heart disease) on mortality in stroke were trained with 6 different MLP algorithms [quick propagation (QP), Levenberg-Marquardt (LM), backpropagation (BP), quasi-Newton (QN), delta bar delta (DBD), and conjugate gradient descent (CGD)]. The performances of the MLP neural network algorithms were compared using the receiver operating characteristic (ROC) curve method. Results: Among the 6 algorithms that were trained with the MLP, QP achieved the highest specificity (81.3%), sensitivity (78.4%), accuracy (80.7%), and area under the curve (AUC) (0.869) values, while CGD achieved the lowest specificity (61.5%), sensitivity (58.7%), accuracy (60.8%), and AUC (0.636) values. The AUC of the QP algorithm was statistically significantly higher than the AUCs of the QN, DBD, and CGD algorithms (P < 0.05 for all of the pairwise comparisons). Conclusion: The MLP trained with the QP algorithm achieved the highest specificity, sensitivity, accuracy, and AUC values. This can be helpful in the prediction of mortality in stroke.

Anahtar Kelimeler:

Key words: Multilayer perceptron neural networks, stroke, mortality, algorithm

Prediction of mortality in stroke patients using multilayer perceptron neural networks

Keywords:

Key words: Multilayer perceptron neural networks, stroke, mortality, algorithm,

PDF

___

American Stroke Association. About stroke. Dallas (TX): ASA. Available from URL: http://www.strokeassociation.org/ STROKEORG/AboutStroke/About-Stroke_UCM_308529_ SubHomePage.jsp. Accessed 17 January 2011. 2. Yan H, Jiang Y, Zheng J, Peng C, Li Q. A multilayer perceptron- based medical decision support system for heart disease diagnosis. Expert Syst Appl 2006; 30: 272-81.
Türe M, Kurt İ, Kurum AT, Özdamar K. Comparing classifi cation techniques for predicting essential hypertension. Expert Syst Appl 2005; 29: 583-8. 9. Marcos JV, Hornero R, Álvarez D, Campo F, Zamarrón C, López M. Utility of multilayer perceptron neural network classifi ers in the diagnosis of obstructive sleep apnoea syndrome from nocturnal oximetry. Comput Meth Prog Bio 2008; 92: 79-89.
İçer S, Kara S, Güven A. Comparison of multilayer perceptron training algorithms for portal venous doppler signals in the cirrhosis disease. Expert Syst Appl 2006; 31: 406-13.
Güler I, Übeyli ED. An expert system for detection of electrocardiographic changes in patients with partial epilepsy using wavelet-based neural networks. Expert Syst 2005; 22: 62- 71.
Santhosh Baboo S, Kadar Shereef I. An effi cient weather forecasting system using artifi cial neural network. IJESD 2010; 1: 321-6.
Rumelhart DE, McClelland J, PDP Research Group. Parallel distributed processing: explorations in the microstructure of cognition. Cambridge (MA): MIT Press; 1986.
Bishop CM. Neural networks for pattern recognition. Oxford: Oxford University Press; 1995.
StatSoft . StatSoft electronic textbook. Available from URL: http://www.statsoft .com/textbook/stathome.html. Accessed 15 January 2011.
Güler NF, Übeyli ED, Güler İ. Recurrent neural networks employing L yapunov exponents for EGG signals classifi cation. Expert Syst Appl 2005; 29: 506-14.
Rumelhart D, Hinton G, Williams R. Learning internal representations by back-propagating errors. Cambridge (MA): MIT Press; 1986.
Duda RO, Hart PE, Stork DG. Pattern classifi cation. New York: Wiley, 2001.