Banca de QUALIFICAÇÃO: MARCUS VINICIUS DE CASTRO OLIVEIRA

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : MARCUS VINICIUS DE CASTRO OLIVEIRA
DATE: 23/06/2023
TIME: 08:00
LOCAL: Ambiente virtual
TITLE:

Low computational cost automatic anatomical features extraction from eye fundus images for glaucoma diagnosis.

KEY WORDS:

 Image Processing, Machine Learning, Feature Extraction, Glaucoma.

PAGES: 50
BIG AREA: Ciências Exatas e da Terra
AREA: Ciência da Computação
SUMMARY:

Glaucoma is a disease that progressively affects the optic nerve and is the leading cause of blindness worldwide. One of the most accurate strategies for diagnosis is clinical analysis aided by results from Optical Coherence Tomography (OCT). The OCT can identify anomalies in the anatomy of the optic nerve, providing support for medical diagnosis. However, this examination comes with a high cost, which tends to hinder its widespread use. As an alternative to this approach, basically two lines of research are proposed in the literature: 1) the use of deep neural networks, and 2) solutions that use processing of retinal fundus images. The problem with the first approach is its high computational cost and the need for large volumes of training data (which are not publicly available). Regarding the second approach, due to the differences in data sets - whether in terms of tonality, image size, resolution, among others - the proposed methods show satisfactory results only on the specific dataset they were trained on. In this context, aiming to address the aforementioned limitations, the objective of this work is to develop a method that can standardize the extraction of optic nerve anatomy characteristics through processing of retinal fundus images in various available datasets. Subsequently, after the extraction of these characteristics, a classifier based on a machine learning algorithm will be used. As a result, a low computational cost system is expected to assist in the diagnosis of glaucoma.

BANKING MEMBERS:
Interno - 1674068 - LEONARDO CHAVES DUTRA DA ROCHA
Presidente - 1758759 - MICHEL CARLO RODRIGUES LELES