Spectral-morphological analysis of acoustical images of biological tissues and composite structures: I. Statistical approach

摘要

The problem of classifying images of different biological tissues and composite structures is solved using the spectral and morphological analysis based on the Bayesian method for statistical hypothesis verification. The basis functions are constructed from a learning set. The spectral approach and its particular realizations in the form of Bartlett's and Pisarenko's methods adapted to the problem are considered. An extension of the spectral approach to the more general spectral-morphological classification is proposed. The latter takes into account the spatial-spectrum features of the structure types to be classified, as well as their morphological features, which manifest themselves in a correlation between the expansion coefficients. The characteristic properties of the spectral and spectral-morphological approaches are discussed using numerical classification examples. The method is generalized to the classification of multiparameter images of structures, which may be represented, for example, by the distributions of the sound velocity, density, absorption, and values of the nonlinear parameter. (C) 2005 Pleiades Publishing, Inc.