Neural Network Based Virtual Reality Spaces for Visual Data Mining of Cancer Data: An Unsupervised Perspective.

Unsupervised neural networks are used for constructing virtual reality spaces for visual data mining of gene expression cancer data. Datasets representative of three of the most important types of cancer considered in modern medicine (liver, lung and stomach) are considered in the study. They are composed of samples from normal and tumor tissues, described in terms of tens of thousands of variables, which are the corresponding gene expression intensities measured in microarray experiments. Despite the very high dimensionality of the studied patterns, high quality visual representations in the form of structure-preserving virtual spaces are obtained using SAMANN neural networks, which enables the differentiation of cancerous and noncancerous tissues. The same networks could be used as nonlinear feature generators in a preprocessing step for other data mining procedures. [ABSTRACT FROM AUTHOR]