engleski

Rational Variety Mapping for Contrast-Enhanced Nonlinear Unsupervised Segmentation of Multispectral Images of Unstained Specimen

A methodology is proposed for nonlinear contrast-enhanced unsupervised segmentation of multispectral (color) microscope image of principally unstained specimen. The methodology exploits spectral diversity and spatial sparseness to find anatomical differences between materials (cells, nuclei and background) present in the image. It consists of the rth-order rational variety mapping (RVM) followed by matrix/tensor factorization. Sparseness constraint implies duality between nonlinear unsupervised segmentation and multi-class pattern assignment problems. Classes not linearly separable in the original input space become separable with high probability in the higher-dimensional mapped space. Hence, RVM mapping brings twofold effect: it takes implicitly into account nonlinearities present in the image (they are not required to be known) and increases spectral diversity (contrast) between materials that occurs due to increased dimensionality of mapped space. This is expected to improve performance of systems for automatic classification and analysis of microscope histopathological images. The methodology is validated using RVM of the 2nd- and 3rd-orders of the experimental multispectral microscope images of the unstained nerve fibers (n. ischiadicus) and the unstained white pulp in the spleen tissue by comparison with a manually defined ground truth labeled by two trained pathophysiologists. The methodology can be useful for additional contrast enhancement of the image of stained specimen as well.

pathology; (un)staining; multispectral image; nonlinear segmentation

nije evidentirano