On the relevance of nonlinear analysis of time series associated to CT kidney images

The goal of this paper is to present a set of algorithms — based on nonlinear time series analysis methods, use them for discrimination between normal and modified kidney tissue and conclude if this analysis is trustworthy. A set of 100 CT images of normal and benign/malign affected renal tissue was used. The classification procedure implies the association of a time series to every CT image and computation of the correlation dimensions of the reconstructed attractors over a range of embedding spaces. The paper concludes that the properties of the associated time series can be used to discriminate between normal and modified tissue. This classification tool can be utilized as diagnostic confirmation support and the results can be improved by a further added database.