Image Analysis and Machine Learning for OCT Image Sequences.

The eye provides a window through which structural and functional characteristics of blood vessels can be visualised and quantified using new technology. This has potential impacts for a wide range of clinical medical problems. This project will develop computational tools for data analysis so that clinically actionable quantitative information can be derived from retinal images. Optical coherence tomography (OCT) is a commonly used technique for capturing three-dimensional volumetric measurements of the retina. Manual segmentation of the edges in OCT images to detect retinal layers is time-consuming work. There has been some success in using deep learning techniques with such images, however such approaches rely on large amounts of training data and have not been used to segment the full set of layers present in the image. In this project we propose to develop image processing and machine learning techniques to build computational models capable of quickly processing an input image, and providing meaningful layering information about the image, including maps of layer thicknesses and volumes across the macula. These techniques substantially benefit from relying on the known underlying structure of the image available via image processing, which does not need a large training set of labelled images to work effectively.