Robust deep learning-based approach for retinal layer segmentation in optical coherence tomography images

Abstract

[Abstract]: In recent years, the medical image analysis field has experienced remarkable growth. Advances in computational power have made it possible to create increasingly complex diagnostic support systems based on deep learning. In ophthalmology, optical coherence tomography (OCT) enables the capture of highly detailed images of the retinal morphology, being the reference technology for the analysis of relevant ocular structures. This paper proposes a new methodology for the automatic segmentation of the main retinal layers using OCT images. The system provides a useful tool that facilitates the clinical evaluation of key ocular structures, such as the choroid, vitreous humour or inner retinal layers, as potential computational biomarkers for the analysis of different neurodegenerative disorders, including multiple sclerosis and Alzheimer’s disease.