Fixation stability readily obtained from confocal color fundus imaging

Abstract

Purpose: Stabile fixation underpins most visual tasks such as reading, and is important for accurate assessment of visual function and treatment. Retinal imaging instruments average images over time to improve the signal to noise ratio, discarding useful eye movement data. We determined whether the frame-to-frame motion of the retina during non-mydriatic color fundus imaging could provide fixation stability measures, e.g. Bivariate Contour Ellipse Area (BCEA).

Methods: Non-mydriatic color fundus images were acquired using the Digital Light Ophthalmoscope (DLO). Twenty subjects with varied fundus pigmentation were tested without mydriasis. The DLO uses a digital light projector with LED light sources to provide the illumination for both confocal imaging and fixation stimuli. The DLO projects a series of lines across the fundus that is synchronized to the 2D CMOS sensor, providing high contrast confocal imaging. Monochromatic 40 deg images were acquired with alternating red and green LED illumination at 14.3 Hz and overlayed to present a pseudo-color image to the operator in real time. To reduce pupil constriction and patient discomfort, the green LED was long-pass filtered with a 570 nm filter. A 1.5mm entrance pupil and time-averaged power of <30 uW were used. Images were aligned automatically with custom software (MATLAB) using cross-correlation and 2D translation. A canvas of twice the image size was used to allow image alignment despite moderate eye movements. Blinks and large saccades were discarded and BCEA was computed.

Results: The image alignment algorithm successfully aligned nearly all the frames, rejecting 3.7%, and allowing fixation stability to be computed from color fundus image data. The BCEA for 1 standard deviation was 2.97 log minarc² for all subjects and both the red and yellow-orange illumination. There was no difference between the BCEA for red or yellow-orange illumination (t = .86).

Conclusions: The color DLO records sufficiently high quality images to reliably calculate measures of fixation stability. Despite recruiting an especially challenging population that included dark fundi, small pupils, high refractive errors, and media issues, we achieved success in all subjects tested to date.