The usefulness of two time-series modeling techniques (autoregressive (AR) modeling and complex autoregressive (CAR) modeling) are investigated for shape description of substructures in CT images. For this purpose, the organ to be identified is separated from the section of a CT image by applying edge detection followed by edge linking, and the boundary of the substructure is extracted in terms of a sequence of contour coordinates that can be viewed as a time series. The modeling techniques are then applied to obtain AR and CAR coefficients. These coefficients serve as feature vectors that represent shapes of substructure boundaries. The feature vectors of known substructure and of unknown substructure obtained by the same method are compared and their identity ascertained by a matching technique involving computation of Euclidean distances. As CT images are invaluable in abdominal investigations, images with the liver as the central organ have been used to check the efficacy of the models.