Processing of adaptive optics photoreceptor images and application to the study of healthy and diabetic retinas

The introduction of adaptive optics (AO) into vision science has made it possible for clinicians to study the human retina in vivo with high-resolution images. The study of AO images of the cone photoreceptor mosaic is becoming a fundamental step in the assessment and monitoring of the health of the retina and in the understanding of the photoreceptor physiology. However, the development of automated algorithms for the analysis of such high amount of information is a necessary step towards the use of AO imaging in clinical practice. In this thesis we aimed at developing a procedure for the automated analysis of the cone mosaic in AO images and we showed its application to the study of properties of cone reflectance in healthy eyes and eyes affected by retinopathy. The work towards the achievement of this aim is presented here in the form of three journal publications and one conference paper. Using a custom developed technique for the simulation of realistic cone images, we optimised and evaluated the performance of automated cone detection algorithms. Using automated cone detection and semi-automated retinal vessel segmentation, we analysed the cone mosaic of a healthy subject over time. We observed that the difference in cone reflectance increases with the time separation between the data acquisitions, negatively affecting the tracking of the same cones over time. With the same method we then investigated cone reflectance in healthy and mild non-proliferative diabetic retinopathy subjects. We were able to determine cone reflectance metrics that quantified reflectance spatial distribution and showed a significant difference between the two groups. Finally, we discuss possible future directions for research that could build on our results