Automated Analysis of OCT Images of the Crystalline Lens

摘要

Presbyopia is the age related, gradual loss of accommodation, mainly due to changes in the crystalline lens. As part of research efforts to understand and cure this condition, ex vivo, cross-sectional OCT images of crystalline lenses were obtained and analyzed to extract their physical and optical properties. The raw OCT images are distorted, as the probing beam passing through media of different refractive indices and refraction on curved surfaces. In a first step, various filters, edge detection and pattern matching methods are applied to isolate the edge contour. An ellipse is fitted to the lens outline to obtain central reference point for transforming the pixel data into the analysis coordinate system. This allows for the fitting of high order equation to obtain a mathematical description of the edge contour, which obeys constraints of continuity as well as zero to infinite surface slopes from apex to equator. Robustness of these algorithms are tested by analyzing the images at various contrast levels. Gradient refractive index of the lens is determined and the physical shape is reconstructed. In a further refinement, the refraction on the curved anterior surface is compensated to obtain the actual shape of the posterior surface. Once the physical shape is fully reconstructed, the optical properties are determined by fitting conic sections to both surfaces and calculating the power profile across the lens. The relative contribution of each of these refinement steps is investigated by comparing their influence on the effective power of the lens.