Diagnóstico de glaucoma em imagens de fundo de olho utilizando estatística espacial

• Ano de defesa: 2018
• Autor(a) principal: SILVA, Camila Costa
• Orientador(a): PAIVA, Anselmo Cardoso de
• Banca de defesa: BORCHARTT, Tiago Bonini , AIRES, Kelson Romulo Teixeira
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Maranhão
• Programa de Pós-Graduação: PROGRAMA DE PÓS-GRADUAÇÃO EM CIÊNCIA DA COMPUTAÇÃO/CCET
• Departamento: DEPARTAMENTO DE INFORMÁTICA/CCET
• País: Brasil
• Palavras-chave em Português: Retinografia; Máquina de Vetores de Suporte; Estatística Espacial
• Palavras-chave em Inglês: Glaucoma; Retinography; Support Vector Machine; Spatial Statistics
• Área do conhecimento CNPq: Processamento Gráfico
• Link de acesso: https://tedebc.ufma.br/jspui/handle/tede/2255

Glaucoma is a disease of the retina considered the second leading cause of blindness in the world reaching an approximate prevalence between 1% and 2% of the population, according to data from the World Health Organization (WHO). It is usually caused by increased fluid pressure in the optic nerve, leading to progressive and irreversible loss of vision. Early diagnosis of glaucoma is therefore critical to ensure a prompt and adequate treatment, being able to minimize damage and prevent loss of vision. Thus, the use of detection and diagnostic systems (CAD - Computer Aided Detection and CADx - Computer Aided Diagnosis) to assist the specialist has increased the chances of correct diagnoses. Photographs of fundus eye (retinographies) are used as a valuable resource in medical diagnoses, as they often present indications about eye diseases such as glaucoma. In this context, this work presents a methodology based on image processing to diagnose glaucoma from the image texture analysis represented using Compound Local Binary Pattern and spatial statistics applied in medical images of retinographies. The method is applied on the region of interest that represents the optical disk, whose segmentation is based on the gold standards in the RIM-ONE public base. Samples are preprocessed using the opponent colors method followed by histogram equalization. The Moran Index and Ripley’s K function are used to describe the texture of the optical disc region. The Support Vector Machine is used to perform classification. The method presented promising results, reaching 95.08% accuracy, 93.4% sensitivity and 96.4% specificity as the best result, and it demonstrates a remarkable performance of the proposed methodology when compared to the methods available in the literature for diagnosis of glaucoma.