Orientador: Yuzo Iano / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Elétrica e de Computação / Made available in DSpace on 2018-08-21T13:38:13Z (GMT). No. of bitstreams: 1 Arthur_AngelicaMoises_M.pdf: 3041531 bytes, checksum: 853fd2e77e1d464c53c93910d6eed4bc (MD5) Previous issue date: 2012 / Resumo: Este trabalho de Mestrado visa a segmentação de imagens de alta resolução da retina humana para favorecer o treinamento de bases de dados e facilitar a identificação automática de patologias. O uso de novas tecnologias na área da oftalmologia vem crescendo à medida que novos equipamentos, que utilizam imagens digitais e que possuem alto poder computacional, têm sido desenvolvidos. Neste trabalho, buscou-se processar as imagens captadas por um novo retinógrafo denominado Retinal Function Imager (RFI), adquirido por meio de um projeto de pesquisa temático FAPESP da FCM (Faculdade de Ciências Médicas) em parceria com a FEEC (Faculdade de Engenharia Elétrica e de Computação), ambas da Unicamp. A microvasculatura da retina humana possui uma única e importante característica de permitir a visualização de parte da circulação humana de forma direta, não invasiva e in vivo, prontamente fotografada e sujeita a análise de imagem digital. A partir disso, a análise automática de diversas patologias, como retinopatia diabética e degeneração macular, tem sido proposta em diversos artigos da literatura nas últimas duas décadas, dispensando muitas vezes o uso de um especialista observador e diminuindo os custos desse diagnóstico. Além disso, medidas quantitativas da topografia vascular da retina usando análise de imagem digital a partir da fotografia da retina, têm sido utilizadas como ferramentas de pesquisa para melhor se compreender a relação entre a microvasculatura da retina e doenças cardiovasculares. O RFI aqui utilizado permite a visualização de estruturas até então não acessíveis. A partir das imagens captadas, foram propostas técnicas para se detectar automaticamente características marcantes do fundo de olho, tais como disco óptico, região foveal e topologia vascular. Por fim, são realizadas segmentações nas imagens para facilitação de treinamento de bases de dados e do diagnóstico automático de patologias. A partir dos testes realizados, 100% das imagens testadas tiveram a região do disco óptico localizada corretamente e em 92% houve sucesso na segmentação da região foveal. Os resultados encontrados mostram que os algoritmos propostos podem ter papel importante para a obtenção de medidas quantitativas da retina humana e também ajudar em pesquisas que buscam a relação entre mudanças de doenças cardiovasculares sistêmicas e vasculares da retina / Abstract: This work aims to segmentation of high resolution human retina images to promote the training database and facilitating automated identification of diseases. The use of new technologies in the field of ophthalmology is growing as new equipment, which use digital images and have high computational power, has been developed. In this work, we process the images captured by a new retinal camera called Retinal Function Imager (RFI), acquired from a FAPESP research project from FCM (Faculty of Medical Sciences) in partnership with FEEC (Faculty of Electrical Engineering and Computer), both of Unicamp. The microvasculature of the human retina has a unique and important feature to allow viewing of the human circulation in a direct, non-invasive, and in vivo way, readily photographed and subjected to digital image analysis. From this, the automatic analysis of several pathologies such as diabetic retinopathy and macular degeneration has been proposed in several articles in the literature over the past two decades, often eliminating the use of an expert observer and reducing the costs of this diagnosis. Moreover, quantitative measurements of the topography retinal vascular using digital image analysis from the retina photograph have been used as research tools to better understand the relationship between micro vascular retinal and cardiovascular diseases. The RFI used in this work allows the visualization of structures not previously accessible. From the captured images are proposed techniques to automatically detect salient features of the optic disc, vessel topology, and fovea. Finally, segmentations are performed on the images to facilitate the training database and the automatic diagnosis of pathologies. From the tests, the optic disc regions were correctly located in 100% of the tested images and the foveal regions were correctly segmented in 92% of the tested images. The results show that the proposed algorithms can be important to obtain quantitative measures of the human retina and also help in research seeking the relationship between changes in systemic cardiovascular and retinal vascular diseases / Mestrado / Telecomunicações e Telemática / Mestra em Engenharia Elétrica

Indiana University-Purdue University Indianapolis (IUPUI) / Down syndrome (DS) is a genetic condition caused by the triplication of human chromosome 21 and presents with many phenotypes including decreased brain size, hypocellularity in the brain, and assorted ocular phenotypes. Some of the ocular phenotypes seen are increased risk of cataracts, accommodation difficulties, increased risk of refractive errors, and increased retinal thickness. The Ts65Dn mouse model of DS is a classically used mouse model as it presents a number of phenotypes also seen in those with DS. Some of these phenotypes include decreased brain volume, abnormal synaptic plasticity, and ocular phenotypes. These ocular phenotypes include decreased visual acuity, cataracts, and increased retinal thickness. The Ts65Dn mouse model is trisomic for Dyrk1a, a gene of interest in DS research. We hypothesize that there will be a genotypic and sex effect of retinal thickness and retinal ganglion cell (RGC) count at postnatal day 15 in the Ts65Dn mouse model of DS. Retinal slices were taken from male and female trisomic and euploid Ts65Dn mice at P15 and fluorescently labeled for RGCs and bipolar cells via immunohistochemistry. The retinas were measured for total retinal thickness and RNA-binding protein (RBPMS) positive cells in the RGC layer were counted. There was no genotypic or sex effect when comparing retinal thickness in trisomic mice as compared to euploid mice. There was a genotypic effect of RBPMS positive cell count in which the trisomic mice had a higher number of RBPMS positive cells than euploid mice. Increased retinal thickness along with increased RGC number have both been implicated with decreased apoptosis in the retina. In the Ts65Dn mouse model along with in individuals with DS, this could be due to an increase in DYRK1A protein levels reducing apoptosis. In future studies, determining DYRK1A’s influence in retinal thickness and RGC number could result in a treatment for overactive DYRK1A that could normalize retinal thickness and RGC number in those with DS.

Typescript (photocopy). / Digitized by Kansas Correctional Industries

published_or_final_version / Anatomy / Master / Master of Medical Sciences

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.