Abstract
Purpose :
Many different instruments record an image of the eye, and process these images to find the pupil boundary, outer iris boundary, structure and other features. These features may be used to guide surgery, establish registration between instruments, identify the patient, and determine fixation. However, there is a lot of variation between different demographic groups in the appearance of the eye image, and the instrument illumination and camera characteristics also affect the image data. Since it is often desirable to establish illumination settings that optimize across the population, a method is needed to find the correct eye illumination settings automatically.
Methods :
An example instrument which has software controllable illumination at 780 and 940 nm was used to systematically vary the illumination levels while synchronously recording the eye images. By recording data rapidly, a full sequence of images could be acquired in a few seconds. Data was acquired for various demographic groups, and the images were analyzed with automatic analysis software to determine the range of settings that would provide optimum results.
Results :
With 10 different settings for 940 nm and 10 settings for 780 nm, it was found that for any given eye, between 25 and 40 of the 100 illumination combinations (dependent on eye color) would provide images that would meet a given success criteria. The different racial groups had different ranges, but the ranges overlapped in most areas. However, it was difficult to identify a single combination of settings that would work for all individuals in all groups. For example, the settings that worked best on light blue eyes were quite different from those that worked on darker Asian eyes.
Conclusions :
While a single combination of illumination settings would provide adequate images for most groups, it was difficult to find a single combination of settings that would work for all individuals. This implies that there would need to be some operator control of the illumination settings, or some selection based on eye color or other demographic.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.