Sistema para montagem de cariótipo de peixe baseado em processamento digital de imagens

• Ano de defesa: 2018
• Autor(a) principal: Goedert, Matheus de Lima
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Medianeira; Brasil; Programa de Pós-Graduação em Tecnologias Computacionais para o Agronegócio; UTFPR
• Palavras-chave em Português: Cromossomos; Citogenética; Processamento de imagens - Técnicas digitais; Chromosomes; Cytogenetics; Image processing - Digital techniques; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO; Engenharia/Tecnologia/Gestão
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/3448

By the principle that evolutionary explanations are usually genetic explanations, obtaining information about the species genome is fundamental in the validation process of some evolutionary theories. In this kind of research, the cytogeneticists use some tools, such as the assembly of karyotypes and their representation, which allows the researcher to identify morphological differences and similarities in the chromosomes, in a relatively simple way. However, this task can be kind of laborious and exhaustive. Therefore, many computational systems involving Digital Image Processing (DIP) have been recommended with the purpose of assisting in this activity. It is notable that DIP presents itself as an important tool for the processes automation, either by its agility, precision or comfort. Thereby, this work aims to develop a system based on the concepts of computer vision for automatic classification of fish chromosomes from individual metaphases images. In order to do this, it was used DIP techniques implemented in C ++ through the OpenCV library. The software used light field microscopy images of mitotic metaphases from individual cells of the Hoplias malabaricus (traíra) fish, collected from the Molecular and Chromosome Biodiversity Laboratory database of UTFPR Santa Helena campus. Segmentation was the first stage of image processing, and it has been divided into background elimination, first segmentation and second segmentation. This step’s fundamentals were the automatic image thresholding, eliminating its background, and the separation of joined chromosomes, using a k-means clustering algorithm. Due to the colors and shades diversity of its respective backgrounds, this step can be considered the most challenging and consequently the most extensive and computationally costly. Ultimately, the chromosomes individualized images were classified according to Levan (1964) proposal. An algorithm was developed in order to scan the chromosome in the searching of centromere position and posteriorly to measure the arms. The segmentation results were satisfactory, presenting 91.25% efficiency. However, the classification showed a lower efficiency, around 63.64%, giving the system an overall efficiency of approximately 60.00%.