Automating the eye examination using optical coherence tomography

Optical coherence tomography (OCT) devices are becoming ubiquitous in eye clinics worldwide to aid the diagnosis and monitoring of eye disease. Much of this uptake relates to the ability to non-invasively capture micron-resolution images, enabling objective and quantitative data to be obtained from ocular structures. Although safe and reasonably quick to perform, the costs involved with operating OCT devices are not trivial, and the requirement for OCT and other imaging in addition to other clinical measures is placing increasing demand on ophthalmology clinics, contributing to fragmented patient pathways and often extended waiting times. In this thesis, a novel “binocular optical coherence tomography” system that seeks to overcome some of the limitations of current commercial OCT systems, is clinically evaluated. This device incorporates many aspects of the eye examination into a single patient-operated instrument, and aims to improve the efficiency and quality of eye care while reducing the overall labour and equipment costs. A progressive framework of testing is followed that includes human factors and usability testing, followed by early stage diagnostic studies to assess the agreement, repeatability, and reproducibility of individual diagnostic features. Health economics analysis of the retinal therapy clinic is used to model cost effectiveness of current practice and with binocular OCT implementation. The binocular OCT and development of other low-cost OCT systems may improve accessibility, however there remains a relative shortage of experts to interpret the images. Artificial intelligence (AI) is likely to play a role in rapid and automated image classification. This thesis explores the application of AI within retinal therapy clinics to predict the onset of exudative age-related macular degeneration in fellow eyes of patients undergoing treatment in their first eye. Together with automated and simultaneous imaging of both eyes with binocular OCT and the potential for low-cost patient-facing systems, AI is likely to have a role in personalising management plans, especially in a future where preventive treatments are available