Assessing diabetic retinopathy by means of lesions detection in eye-fundus images

Identifier:  TFG:6954

Handle: https://hdl.handle.net/20.500.11797/TFG6954

Authors:  Iglesias Burgos, Alberto

Abstract:
Diabetic retinopathy, a major cause of preventable vision loss, primarily affects the global working-age population. This project explores using computer-aided diagnostics to assess diabetic retinopathy severity by analyzing images with various lesions, including microaneurysms and hemorrhages. Multiple public retinal image datasets undergo automatic analysis via deep learning, specifically the LezioSeg model. The study combines human and AI-based image analysis to quantify lesions, evaluate diabetic retinopathy severity, and compare results with ophthalmologists. Initial results fell short, leading to careful curation of a new dataset with Hospital Universitario Sant Joan de Reus to enhance the LezioSeg model and address initial analysis challenges.