Abstract
Purpose :
We examined the ability of an artificial intelligence algorithm, iKey®, to assist with automatic identification of silent optic nerve head changes in a diabetic population. iKey® has been shown to be 91% sensitive for detection of vascular pattern change on optic nerve heads (changed fig 1). We set out to analyse images from a referral pool of patients with potential glaucoma and to develop an algorithm to identify the location of change as alerted to by the iKey algorithm.
Methods :
Images where change of blood vessel feature vectors was alerted by iKey® were then processed as follows by a novel change localisation module: the module, given two images of the same eye, finds the optic disc and crops them. Both cropped images are aligned using histogram equalisation filters and ORB. The blood vessels are segmented on the cropped images and the superimposed blood vessels are displayed (fig 2). Similar areas are shown in Gray and changed loci in colour. The blood vessel segmentation is superimposed by using the three different channels of the Red Green Blue images. The first image is represented using the Blue and Green channels, creating light blue. The second image is represented using the Red channel. Pixels that represent blood vessels in both images have all three channels red, green and blue to the maximum value creating a white colour. and image overlay.
Results :
Findings confirmed iKey change alert using localisation of vessel shift on optic nerve heads.
Conclusions :
It is possible to use automatic analysis of retinal photographs to identify areas of change on the optic nerve head images. Further research will examine the use of change -detection algorithms for optimising the referral strategies for diabetic patients to busy eye centers, in particular query early glaucoma, for further examination.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.