Histopathology image classification via characterization of nuclei arrangement

Abstract. The automatic characterization of histopathology images is an important requirement for the development of computarized tools that might benefit clinicians in their everyday professional workflow. Researchers have developed image descriptors for histopathology images that are mainly migrat...

Full description

Autores:: Álvarez Corrales, Pablo Arturo

Tipo de recurso:

Fecha de publicación:: 2017

Institución:: Universidad Nacional de Colombia

Repositorio:: Universidad Nacional de Colombia

Idioma:: spa

Description
Summary:	Abstract. The automatic characterization of histopathology images is an important requirement for the development of computarized tools that might benefit clinicians in their everyday professional workflow. Researchers have developed image descriptors for histopathology images that are mainly migrated from the techniques used with natural images, which result in high-dimensional feature vectors that are difficult to interpret. Since automatic analysis of histopathology images is performed only as support tools for physicians, the level of interpretability of such automatic analysis is of considerable importance. This thesis work was focused in finding a way towards the characterization of histopathology images through the abstraction of simple histological concepts, i.e. information from cell properties. More specifically, we have investigated the potential of cells' area and topology for the construction of descriptors for histopathology images. Experimental results suggest that our proposed descriptors provide a discriminative power that could be used either for classification or retrieval tasks.

Histopathology image classification via characterization of nuclei arrangement

Publicaciones similares