A NONLINEAR STUDY OF TIME SERIES ASSOCIATED TO CT LIVER IMAGES

This paper is the result of a study of 50 CT images of liver tissues of 10 subjects. The investigations were performed on both normal and cholangiocarcinoma cases. First, applications to extract one dimensional data (time) series, to reconstruct the attractor in the embedding space and to compute its correlation dimension are developed. In the second part, these tools are applied to classify normal areas (benign changes) versus anomalies (malign changes) in the CT livers images. By using statistical tests, we get the conclusions that the connectedness of the attractor and correlation dimensions can discriminate between benign and malign changes.