Reflex Free Fundus Imaging of Wide Field of Views through Small Pupils using DMD Projector Illumination

Abstract

Fundus cameras illuminate and view the retina through the eye pupil. Reflections and haze from the eye and the lenses in the common detection and illumination path obscure the image and limit the field of view and minimum pupil size. Pupil splitting was first demonstrated by Allvar Gullstrand in the earlier 20th century to keep the illumination and detection paths separate throughout the eye to avoid reflexes and haze. In this dissertation, first order optics and nonsequential raytracing were used to determine the size, shape, and locations of the reflexes in a pupil splitting fundus camera, to show how the pupil size and field of view are limited when using full field illumination, and finally to determine how the reflexes are affected by varying the illumination pattern at the pupil and retinal conjugate planes.Varying the irradiance pattern at the exit pupil between successive images while illuminating the entire retina shifts the reflex positions on the sensor while the retinal image remains stationary, so that it is possible to combine two or more images for a reflex free image. Conversely, scanning the retina with bands of illumination in the direction of the pupil separation through a fixed exit pupil keeps the illumination and detection separate through the anterior segment and the Crystalline lens and creates an offset between the area of interest and the reflex so that the reflex appears to come from a different part of the retina than that being illuminated. The conclusion is that to image large fields of views through small pupils, it is better to illuminate and view the retina in sections versus combining two or more images that were each captured with a different irradiance pattern at the exit pupil. To demonstrate the advantages of scanning, a pupil splitting fundus camera was assembled that used a DMD projector to scan illumination stripes across the test eye retina and trigger a camera viewing the retina through the opposite side of the pupil. The pupil split was a simple 3D printed part that held the detection stop, an imaging lens made from back to back achromats that was telecentric in image space, and a fold mirror laterally offset from the detection stop that was used to fold in the illumination. Various relay lens configurations were used to relay the pupil split to the eye pupil including two afocal relays with different magnifications and a single aspheric objective lens. The only moving parts were the sensor that moved towards and away from the imaging lenses for focus and refraction correction and the micromirrors of the DMD. For each projected stripe a full frame was captured and the pixels out of the area of interest were zeroed before assembling the stripes to form a reflex free image. Various stripe widths and step sizes were used to minimize image artifacts and acquisition time. The final design used crossed polarizers in the detection and illumination paths to eliminate the specular reflexes from the common path lenses and scanning to avoid the non-specular reflexes and scatter from the cornea and Crystalline lens. Reflex free images were captured on realistic test eyes with pupils as small as 2.5 mm and fields of view from 60 to 90 degrees.