Purchase this article with an account.
Anthony N. Kuo, Cynthia A. Toth, Joseph A. Izatt; Spatial Correction of Retinal SDOCT Images to Reflect Expected Ocular Curvature. Invest. Ophthalmol. Vis. Sci. 2011;52(14):4050.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Spectral domain optical coherence tomography (SDOCT) is used for high resolution assessment of retinal morphology. SDOCT acquisitions are currently displayed in rectangular form with parallel A-scans. However, this form does not accurately represent true retinal morphology. Relative measurements across this flattened retina (non-axial thicknesses, slopes of surfaces, & others) are not measurements on the real spatial structure. To better reflect actual retinal morphology, we developed algorithms to de-warp retinal SDOCT images.
We developed a simple analytical and a more complex numerical model of the SDOCT retinal scanning geometry. The analytical model assumes a homogenous ocular index of refraction and describes an arc scanning geometry with the scan pivot located at the pupil. The numerical model uses an ‘exact’ model eye in ZEMAX which traces each ray through ocular structures to determine the angle and distance of each retinal image point. Algorithms based on both models were developed to re-map each acquired A scan to its correct location.A healthy volunteer adult subject was imaged with a clinical retinal SDOCT system (Bioptigen). Ocular biometry data from the subject (corneal thickness and curvatures, anterior chamber depth, axial length) was used to inform the models. Conventional B-scan images in rectangular format were compared to de-warped images resulting from the analytical and numerical algorithms, and the retinal radius of curvature from each image type was calculated.
De-warped retinal images resulting from both optical models had a fan-shaped outline and a more natural curved appearance (Figure). Measured retinal radii of curvature from a sample image set were 52.6 mm (conventional image), 16.3 mm (analytical model correction) and 10.0 mm (numerical model correction). Reported model values for the human retina are around 11 mm.
We developed optical models and algorithms to correct retinal SDOCT images to reflect true ocular morphology. Such correction is a pre-requisite for accurate clinical morphometric analyses of the retina.
This PDF is available to Subscribers Only