Semiautomated segmentation of bone marrow biopsies images based on texture features and Generalized Regression Neural Networks

• Autores: Meschino, Gustavo; Moler, Emilce Graciela; Passoni, Lucía Isabel
• Año de publicación: 2004
• Idioma: inglés
• Tipo de recurso: documento de conferencia
• Materia: Ciencias Informáticas; Digital Image Processing; Segmentation; Texture; Classification; Generalized Regression; Neural Networks; Visual; COMPUTER GRAPHICS; Neural nets
• Institución: Universidad Nacional de La Plata

This work presents preliminary results of a method for semi-automatic detection of fat and hematopoietic cells as well as trabecular surfaces in bone marrow biopsies, in order to calculate the percentage of each type of tissue or cell area in relation to the whole area. Experimental results using selected clinical cases are presented. Twenty six biopsies were used, presenting varied distributions of cellularity and trabeculae topography. The approach is based on Digital Image Processing techniques and a Neural Network used for classification using textural features obtained from biopsies images. Results were improved with Mathematical Morphology filters. The algorithm produces highly satisfactory results. The method was shown to be faster and more reproducible than conventional ones, like region growing, edge detection, split and merging. The results from this computer-assisted technique are compared to others obtained by visual inspection by two expert pathologists, and differences of less than 9 % are observed.
Eje: II - Workshop de computación gráfica, imágenes y visualización
Red de Universidades con Carreras en Informática (RedUNCI)