Automated retinal layer segmentation and pre-apoptotic monitoring for three-dimensional optical coherence tomography

The aim of this PhD thesis was to develop segmentation algorithm adapted and optimized to retinal OCT data that will provide objective 3D layer thickness which might be used to improve diagnosis and monitoring of retinal pathologies. Additionally, a 3D stack registration method was produced by modifying an existing algorithm. A related project was to develop a pre-apoptotic retinal monitoring based on the changes in texture parameters of the OCT scans in order to enable treatment before the changes become irreversible; apoptosis refers to the programmed cell death that can occur in retinal tissue and lead to blindness. These issues can be critical for the examination of tissues within the central nervous system. A novel statistical model for segmentation has been created and successfully applied to a large data set. A broad range of future research possibilities into advanced pathologies has been created by the results obtained. A separate model has been created for choroid segmentation located deep in retina, as the appearance of choroid is very different from the top retinal layers. Choroid thickness and structure is an important index of various pathologies (diabetes etc.). As part of the pre-apoptotic monitoring project it was shown that an increase in proportion of apoptotic cells in vitro can be accurately quantified. Moreover, the data obtained indicates a similar increase in neuronal scatter in retinal explants following axotomy (removal of retinas from the eye), suggesting that UHR-OCT can be a novel non-invasive technique for the in vivo assessment of neuronal health. Additionally, an independent project within the computer science department in collaboration with the school of psychology has been successfully carried out, improving analysis of facial dynamics and behaviour transfer between individuals. Also, important improvements to a general signal processing algorithm, dynamic time warping (DTW), have been made, allowing potential application in a broad signal processing field