A smartphone-based proof-of-concept iKey biometric app that verifies optic nerve head (ONH) images for remote diagnosis of eye diseases was recently developed by a team of researchers in Dublin, Ireland. They say this app will help meet the demand for medical personnel in remote areas without access to traditional ophthalmoscopes or fundus cameras.
The iKey captures images of the ONH to use as a routine biometric. “We wanted to develop a software that could bring the color fundus image, with a motive to regularly use and update the image from childhood, to the ubiquitous mobile phone,” the researchers explained. “The more regularly the image is taken, the easier it is to incidentally detect any change to the image which might indicate disease, either with automatic artificial intelligence or telemedicine.”
The researchers chose the small ONH instead of the larger retina for two reasons. “It lies close to the macula, behind the pupil, allowing for easy image capture with even a small pupil for a mobile phone owner, and the ONH contains the uniquely positioned retinal blood vessels and nerve fibers, previously only visible to ophthalmologists, reflecting local and systemic health.”
The app’s verification software for identifying an individual’s unique pattern of retinal blood vessels surpasses that of facial recognition, with 97.06% accuracy. It accumulates a “silent continuous record” of an individual’s ONH images to compare with previously stored images using supervised, specific-feature deep neural networks (DNNs). If an ONH image doesn’t map onto one of the previously enrolled ONH images, verification will fail. This signals to the app user that some asymptomatic change to the ONH has occurred, whether from aging, disease, healing or deterioration.
“The capture of the image may be hampered by media changes, such as cataract and vitreous opacities,” the researchers noted. Otherwise, any specific blood vessel feature changes may signal potentially blinding diseases such as glaucoma or diabetic retinopathy.
In the study, the researchers assessed the platform’s three DNNs, which were developed from anonymous ONH images. The graticule blood vessel (GBV) and blood vessel-specific feature (BVSF) DNNs were trained on unique blood vessel vectors. For comparison, the researchers trained a non-feature-specific (NFS) baseline ResNet50 DNN. They achieved verification accuracies of 97.06% with BVSF, 87.24% with GBV and 79.8% with NFS.
“There’s potential for iKey to be used from an early age, not just for sight protection but also for general health maintenance,” they said. “ONH vessels have been affected by systemic vascular conditions, as well as cerebral small vessel disease and...diseases such as Alzheimer’s. The iKey may have a much broader role for disease prevention and health monitoring, as with other wearable technology. Future work will include the relationship of nonvascular features around the vascular framework within the geometric graticular space.”
Coleman K, Coleman J, Franco-Penya H, et al. A new smartphone-based optic nerve head biometric for verification and change detection. Trans Vis Sci Tech. 2021;10(8):1. [Epub ahead of print].