Application of multilayer perceptron and radial basis function neural networks in differentiating between chronic obstructive pulmonary and congestive heart failure diseases

  • Authors:

  • Saeed Mehrabi;Mehran Maghsoudloo;Hossein Arabalibeik;Rezvan Noormand;Yones Nozari

  • Affiliations:

  • Department of Medical Physics and Biomedical Engineering, Faculty of Medicine, Tehran University of Medical Sciences, Tehran, Iran;Research Center for Science and Technology in Medicine, Tehran University of Medical Sciences, IK Hospital Complex, Keshavarz Boulevard, Tehran 14185-615, Iran;Research Center for Science and Technology in Medicine, Tehran University of Medical Sciences, IK Hospital Complex, Keshavarz Boulevard, Tehran 14185-615, Iran;Faculty of Medicine, Tehran University of Medical Sciences, Tehran, Iran;Faculty of Medicine, Tehran University of Medical Sciences, Tehran, Iran

  • Venue:
  • Expert Systems with Applications: An International Journal
  • Year:
  • 2009

Quantified Score

Hi-index 12.06

Visualization

Abstract

Congestive heart failure and chronic obstructive pulmonary disease have similar symptoms which can make their distinction difficult especially at the time of admission or where the access to echocardiography is limited. The multilayer perceptron (MLP) and radial basis function (RBF) neural networks were used to differentiate between patients (n=266) suffering one of these diseases, using 42 clinical variables which were normalized following consultations with cardiologists. Bayesian regularization was used to improve the generalization of the MLP network. In order to design the RBF network, K-Means clustering was used to select the centers of radial basis functions, k-nearest neighborhood to define the spread and forward selection to select the optimum number of radial basis functions. A 10-fold cross validation was used to assess the generalization procedure. The MLP led to a sensitivity of 83.9%, specificity of 86% and an area under receiver operating characteristic curve (AUC) of 0.889+/-0.02 and RBF network resulted in sensitivity of 81.8%, specificity of 88.4% and AUC of 0.924+/-0.017.