Contributions to Structural and Functional Retinal Imaging via Fourier Domain Optical Coherence Tomography

This thesis encompasses contributions to both structural and functional optical coherence tomography (OCT) retinal imaging. One common theme found throughout this thesis is the generation of synthetic or simulated OCT data in order to quantitatively evaluate various image processing algorithms.;We first design and implement a novel sparsity-based retinal layer segmentation algorithm that identifies the boundaries between retinal layers in OCT structural images. In contrast to most state of the art retinal layer segmentation algorithms, our algorithm does not require overly restrictive a priori assumptions about the expected layer structure and can thus possibly be used to detect unexpected structures that arise in pathological cases. One difficulty with quantitative evaluation of segmentation algorithms applied to in vivo retinal images is that the ground truth segmentation is not known. Motivated by this fact, we design a novel method for generating synthetic structural retinal images for which the ground truth segmentation is known. Our technique improves upon an existing synthetic image generation method both by using a nonparametric smooth representation of the average intensity values in the image and by accurately modeling speckle noise. After verifying that synthetic images generated with our method accurately represent real in vivo retinal images, we show that our segmentation algorithm accurately identifies the retinal layer boundaries in the synthetic images and that our algorithm is robust to image noise.;We next examine functional OCT retinal imaging by studying various methods for measuring blood flow velocities in retinal vessels. We develop a two dimensional simulation of a blood vessel immersed in tissue that mimics functional imaging in a clinical setting more accurately than existing one dimensional simulations of retinal blood flow. This retinal vessel simulation enables quantitative evaluation of the accuracy of Doppler OCT algorithms by providing the ground truth blood flow velocities that Doppler OCT techniques aim to measure. We illustrate the utility of the simulation with two case studies. First, we evaluate the accuracy of two commonly used Doppler OCT algorithms as vessel diameters and associated flow speeds become increasingly small. We show that the phase-resolved Doppler OCT algorithm performs best for large blood vessels while neither algorithm performs well for very small blood vessels and slow blood flow velocities. This is the first time that a quantitative evaluation of the estimation accuracy of the Doppler OCT algorithms is performed using synthetic data with known ground truth blood flow velocities. In the second case study we examine the effect of transverse step size between adjacent axial scans on the phase-resolved Doppler OCT algorithm. We show that the phase-resolved Doppler OCT algorithm systematically underestimates blood flow velocities with increasing transverse step size and that this effect is more pronounced at higher velocities. We propose two techniques for correcting measured velocities that can be used to improve the accuracy of the phase-resolved Doppler OCT algorithm.;As previously mentioned, Doppler OCT techniques are only sensitive to velocities along the direction of the incident beam (axial velocities). While Doppler OCT is commonly used to measure blood flow near the optic disc, its application to flow estimation in the human macula is limited because many of the blood vessels in this region are nearly perpendicular to the incident beam. In order to overcome this limitation, we introduce a novel method for visualizing blood flow called split-spectrum amplitude-decorrelation angiography (SSADA). In contrast to Doppler phase-based methods, SSADA relies on intensity information only and is sensitive to both transverse (perpendicular) and axial flow velocities. After establishing that SSADA can be used to image flow in the human macula, we perform phantom flow experiments to establish a linear relationship between SSADA measurements and preset blood flow velocities that enables velocity quantification and thus blood flow measurement in the human macula with SSADA. (Abstract shortened by UMI.).