Novel deep learning approaches in optical coherence tomography imaging

Abstract

Optical coherence tomography (OCT) is a non-invasive imaging modality widely used in ophthalmology for visualizing retinal structures. In this thesis, deep learning (DL) technology has been deployed to enhance OCT scan quality and depth. While existing OCT-DL applications focus on superficial retinal layers, they neglect the capability to heighten deeper eye structures, such as the choroid, and disregard the valuable phase information from the OCT signal. Despite the vast information contained in OCT data, current DL approaches struggle to extract its full potential. Here, I present our effort in tackling these challenges, introducing novel ways to analyze OCT images and showcasing the potential of our DL models to enhance deep feature visualization and unveil concealed information by exploiting the complete OCT signal. By maximizing the capabilities of OCT imaging, our findings open new avenues for advanced clinical diagnosis, thereby contributing to a deeper understanding of human diseases.