Rastreamento de células em vídeos 3D

• Ano de defesa: 2018
• Autor(a) principal: Sousa, Davy Oliveira Barros
• Orientador(a): Macedo, Hendrik Teixeira
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Programa de Pós-Graduação: Pós-Graduação em Ciência da Computação
• Palavras-chave em Português: Computação; Processamento de imagens; Técnicas digitais; Ciências médicas; Células; Rastreamento de células; Imagens 3D; Biomedicina
• Palavras-chave em Inglês: Cell tracking; 3D images; Image processing; Biomedicine
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://ri.ufs.br/jspui/handle/riufs/10782

Performing cell tracking is important for curing and preventing diseases. This is due to the fact that cellular motility is linked to several cellular processes. However, performing the analysis of the trajectory of a cell is not a simple job, because to store the information of the trajectory it is necessary a great amount of images, mainly in 3D images. Thus, it is necessary to create algorithms that can perform cell tracking in a practical and automatic way. In this work was given a continuation of the work of conclusion of course (TCC) of Sousa (2015) in order to improve the tracking algorithm already implemented and presented in Sousa (2015). To overcome the shortcomings of the algorithm implemented by Sousa (2015), modifications were made to the segmentation, labeling and tracking phases. In addition, methods have been added for the detection of cell division and the entry and exit of new cells of the video. The validation of the algorithm was done through an evaluation software that has routines for both the segmentation and the tracking part. 11 datasets were used for the validation, and the proposed algorithm was able to obtain results for 7 of them. Among the 7 datasets, 3 were 2D and the other 4, 3D. Although the results were not satisfactory for the 2D datasets, the 3Ds obtained satisfactory results during the tracking phase, with averages of accuracy ranging from 90.1% to 94.4%. From the validation, it was realized that, even having obtained satisfactory values, including some of them better than some found in the state of the art, the algorithm still needs improvements.