DEEP LEARNING ALGORITHMS FOR OPTICAL COHERENCE TOMOGRAPHY IMAGES WITH APPLICATIONS IN GLAUCOMA

Abstract

Glaucoma is an irreversible blinding disorder affecting nearly 70 million people worldwide. Although the exact cause of glaucoma remains unknown, the elevated intraocular pressure (IOP) that results in complex 3D biomechanical changes of the optic nerve head tissues (ONH), is indeed a well-known risk factor and considered as the only clinically treatable one. Since the inception of optical coherence tomography (OCT) technology, it has been possible for the in vivo assessment of these morphological changes in both the neural and connective tissues of the ONH. Nevertheless, it has been possible to use only a single neural tissue parameter (retinal nerve fiber layer thickness) in the clinics for the structural assessment of glaucoma. In this thesis, we wish to offer an accurate and simplified glaucoma diagnosis by leveraging on the power of deep learning (DL) to fully exploit the 3D morphological information present in OCT images.