Abstract
Purpose :
Fundus images are routinely stitched together as montages to provide an increased field-of-view (FOV). Due to the difficulty of precisely controlling overlap between images during acquisition, as well as the inherent distortion of imaging the curved retinal surface onto a 2D plane, the angular FOV subtended by the montage is not usually known. Here, we describe an approach to estimate the FOV of ultra-widefield (UWF) fundus image montages using only the image data, the nominal FOV of the imaging system, and the map-projection used to display the montage.
Methods :
UWF fundus image montages, acquired using CLARUSTM 500 (ZEISS, Dublin, CA), were analyzed retrospectively. The constituent images each covered 90° FOV, and were displayed by the instrument software in a stereographic projection, centered on the fovea. Distance (in pixels) from the fovea is therefore proportional to tan(θ/2), where θ is the visual angle. The angular magnification of the eye was assumed to be 17 mm/radian, per the ISO 10940 standard for fundus cameras. By locating the common projection point (the fovea) in each of the constituent images, the bounds of the montage in the projection space were calculated, giving the FOV limits in visual field angle. The derived information was used to map each image pixel to a 3D rendering, based on a simplified model of a generic eye.
Results :
Figure 1A shows an UWF montage, created from stitching together four 90° images. The FOV subtended at the exit pupil of the eye was estimated to be 131° horizontally and 134° vertically (Figure 1B). Figure 1C shows a corresponding 3D rendering of the image data, generated using a simplified spherical eye model.
Conclusions :
A post-hoc method for estimating the FOV subtended by UWF retinal image montages, and generating a corresponding 3D rendering, was presented. Such an approach may be helpful when visualizing UWF images in conjunction with data from other 2D or 3D modalities (e.g., in the planning of retinal surgery). Though stereographic projection is commonly used to display UWF images in commercial fundus imaging systems, the described approach could be applied to any known map projection.
This is a 2021 Imaging in the Eye Conference abstract.