Improving autofocus performance in a widefield fundus imaging system, using peripheral defocus measurements

Abstract

Purpose : Traditionally, fundus camera designs incorporate a large collection aperture to allow for comfortable levels of illumination, which limits the depth-of-field. For widefield systems, maintaining good focus across the full field-of-view (FOV) can be challenging due to the wide range of human retinal curvatures. In this study, we propose an autofocus scheme for a widefield fundus imaging system, aimed at balancing focus quality throughout both the central and peripheral field.

Methods : A widefield slit-scanning ophthalmoscope, CLARUS™ 500 (ZEISS, Dublin, CA), with prototype software was used for imaging the retina over a 90° FOV. During alignment, the displacement on the retina between two illumination beams entering the pupil at different locations was measured and used to infer the best focus setting. In the proposed approach, focus error is estimated for both central and peripheral retinal regions simultaneously (Figure 1A). An aggregated focus setting is then applied to balance focus quality across the 90° FOV. Color fundus images were captured from 12 eyes at a variety of fixation orientations, totaling 22 different imaged retinal fields. Each field was imaged using (i) autofocus based on an estimate of focus error near the center of the FOV, and (ii) the proposed approach. The captured images were graded for overall focus quality by a licensed clinician; the assigned grades were normalized to a range of 1-5 (with 5 signifying the best focus quality).

Results : The mean scores for focus quality were 3.32 ± 0.84 for central autofocus, and 4.05 ± 0.79 for the proposed autofocus strategy. Using the proposed strategy, 16/22 images achieved a grade of 4 or better, compared to 7/22 using central autofocus.

Conclusions : Our results suggest that an autofocus strategy during image acquisition that attempts to compensate peripheral defocus, as well as central focus error, has the potential to improve the overall quality of focus in widefield fundus images.

This abstract was presented at the 2019 ARVO Imaging in the Eye Conference, held in Vancouver, Canada, April 26-27, 2019.

 

Figure 1. (A) The displacement on the retina between two illuminating beams is used to infer focus error at different retinal locations; (B) 90° fundus image of a myopic eye. Using central autofocus, features near the FOV center appear sharp (C), but important detail in the periphery exhibits defocus (D). Using the proposed autofocus scheme, good image sharpness was maintained both centrally (E) and peripherally (F).