Skip to main content
Open Access Publications from the University of California

UC San Diego

UC San Diego Electronic Theses and Dissertations bannerUC San Diego

Deep Learning Based Multimodal Retinal Image Processing


The retina, the light sensitive tissue lining the interior of the eye, is the only part of the central nervous system (CNS) that can be imaged at micron resolution in vivo. Retinal diseases including age-related macular degeneration, diabetes retinopathy, and vascular occlusions are important causes of vision loss and have systemic implications for millions of patients. Retinal imaging is of great significance to diagnosing and monitoring both retinal diseases and systematic diseases that manifest in the retina. A variety of imaging devices have been developed, including color fundus (CF) photography, infrared reflectance (IR), fundus autofluorescence (FAF), dye-based angiography, optical coherence tomography (OCT), and OCT angiography (OCT-A). Each imaging modality is particularly useful for observing certain aspects of the retina, and can be utilized for visualization of specific diseases.

In this dissertation, we propose deep learning based methods for retinal image processing, including multimodal retinal image registration, OCT motion correction, and OCT retinal layer segmentation. We present our established work on a deep learning framework for multimodal retinal image registration, a comprehensive study of the correlation between subjective and objective evaluation metrics for multimodal retinal image registration, convolutional neural networks for correction of axial and coronal motion artifacts in 3D OCT volumes, and joint motion correction and 3D OCT layer segmentation network.

The dissertation not only proposes novel approaches in image processing, enhances the observation of retinal diseases, but will also provide insights on observing systematic diseases through the retina, including diabetes, cardiovascular disease, and preclinical Alzheimer’s Disease. The proposed deep learning based retinal image processing approaches would build a connection between ophthalmology and image processing literature, and the findings may provide a good insight for researchers who investigate retinal image registration, retinal image segmentation and retinal disease detection.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View