Schematic illustration of artificial neural network (ANN) structure. A typical ANN consists of neurons arranged in layers: an input layer (in this case, cDNA microarray expression ratios), one or more hidden layers, and an output layer (in this case, the diagnosis). Data from each layer is processed by mathematical transfer functions to generate the input for the next layer of neurons. During training, data in the output layer is backpropagated through the ANN to correct weights in transfer functions until the ANN achieves the correct output. In this case, expression data from 160 cDNAs selected by gene filtering was input into the ANN. The ANN was educated with a training set consisting of 12 cases of esophageal cancer or Barrett's esophagus, then applied to a test set of 10 new cases. For details, see the article by Xu et al. on page 3493 of this issue.

[Table of Contents]