Deep Learning-based Methods for Extracting Fundus Image Landmarks and Signs of Eye Diseases

Identificador:  TDX:4137

Handle: https://hdl.handle.net/20.500.11797/TDX4137

Autores:  Ali, Mohammed Yousef Salem

Resumen:
Diabetic Retinopathy (DR) is the most typical cause of visual loss in working-age adults. In 2040, it is predicted that over 200 million people will have DR. Glaucoma affects about 75 million individuals worldwide, and it is called the silent thief of sight. Therefore, early diagnosis of glaucoma and DR requires an efficient screening procedure. In addition, medical centers regularly perform eye checkups for patients, especially diabetic ones, to minimize the risk of blindness. Examining eye fundus images is labor-intensive, time-consuming, expensive, and error-prone. Therefore, developing a computer-aided diagnosis (CAD) system that analyzes eye fundus images to help ophthalmologists is crucial. Computer vision techniques, like deep learning and especially convolutional neural networks (CNNs), have significantly improved CAD systems' performance. This thesis uses computer vision techniques to consider four eye tasks: optic disc segmentation, glaucoma detection, segmenting exudate lesions and segmenting other kinds of retinal eye DR lesions from fundus images.