Desenvolvimento de um método para segmentação de imagens histológicas da rede vascular óssea

Abstract

Histology is the area of biology that studies biological tissues. One of the ways to study these tissues is through images. The researcher extracts a sample of an animal, this sample is prepared, sectioned and taken under a microscope, which has a coupled camera, which turns the sample into an image. The analysis of these images is of fundamental importance for the specialists to study and to diagnose possible diseases, malformation or other possible anomalies. One of the tissues that are analyzed in histology is the bone tissue, which is of fundamental importance to protect organs and structure for vertebrate animals. One of the regions analyzed in the bone tissue are the vascular networks of bone which contain bone canals, osteocytes, bone matrix and other artifacts. Bone canals and osteocytes are responsible for the nutrition of the bone tissue. Because of their importance, the specialists study these artifacts in order to discover any damage to these regions and, consequently, to the nutrition of the tissue. Even today the analysis of these artifacts is performed manually by researchers. However, due to the complexity of the histological images, manual analysis takes a lot of time and money from the institutions, and it is a task that depends of the subjective jugments of each evaluator. Literature provides many papers focused on cell nuclei segmentation in histological images, but the automatic segmentation of bone canals and osteocytes is less explored. Due to the lack of research and to assist the specialists in the analysis of the bone vascular network, a method of automatic / semi-automatic segmentation of bone canals and osteocytes is proposed. The method was applied to three diferent image sets which were evaluated through the Dice coeicient and diagnostic approach metrics. In addition to being compared with other automatic methods (neighborhood valley emphasis (NVE), valley emphasis (VE) and Otsu). Results showed that our approach proved to be more eicient than the others, being a viable alternative to analyze the bone vascular network.