Neural networks for automatic diagnosis of labeled monodispersed aerosols in obstructive lung disease

Feed-Forward Artificial Neural Networks (FF-ANN) can be used as classifier when suitably trained. In this paper we trained a FF-ANN to classify images of the lung, in order to develop a system for the automated image-based diagnosis of the bronchial stenosys. After a pre-processing filler we have used an FF-ANN, trained using two backpropagation methods,with momentum and the weight decay, to perform a classification into four classes of disease. The results are very close to the human classification.