International researchers have demonstrated that an artificial intelligence-enabled algorithm can achieve 100% sensitivity in detecting vision-threatening diabetic retinopathy.
US company Eyenuk is behind the autonomous product, the EyeArt AI Eye Screening System, which was the subject of the study which featured in the British Journal of Ophthalmology.
According to the independent research paper, produced by authors from the Moorfields Eye Hospital NHS Foundation Trust, the University College London Institute of Ophthalmology, and the Homerton University Hospital NHS Foundation Trust, the EyeArt system achieved 92-100% sensitivity in diabetic retinopathy detection with multiple retinal imaging platforms.
The study involved 1,257 patients attending annual diabetic eye screening in the UK. The EyeArt System processed images acquired by CenterVue’s EIDON platform with wide-field true-colour confocal scanning technology, as well as images acquired by standard cameras in the English National Diabetic Eye Screening Programme (NDESP).
With EIDON images, EyeArt achieved sensitivities of 92% for any retinopathy, 99% for vision-threatening retinopathy and 100% for proliferative retinopathy. The system’s sensitivities for NDESP images were 92.26% for any retinopathy, 100% for vision-threatening retinopathy and 100% for proliferative retinopathy.
“All processing of the screening episodes was performed by the research team. The vendor was not allowed access to the software or to the data set during the study period,” the authors noted.
“With this [impressive] performance, if the [EyeArt] software were to be hypothetically deployed as a part of the English NDESP, the EyeArt could reduce the need to grade R0M0 (no retinopathy and hence no maculopathy) by half when using EIDON images and by almost two-thirds when using the NDESP images, a considerable workload reduction.”
According to Eyenuk, the UK has been leading the world in diabetic retinopathy screening, achieving patient uptake rates of over 80%, screening nearly 2.5 million diabetes patients annually in England. This is compared with most parts of the world where typically fewer than half of diabetes patients receive annual eye screening.
As a result, diabetic retinopathy is no longer the leading cause of blindness in the working age group in England
“Validation results from the world’s leading diabetic retinopathy screening program once again confirm EyeArt System’s strong diagnostic accuracy and its potential to significantly reduce human workload for DR screening,” said Mr Kaushal Solanki, founder and CEO of Eyenuk.
“We are equally excited to learn that the EyeArt System is shown to be accurate and highly sensitive when used to analyae EIDON images for diabetic retinopathy screening per the English NDESP protocol. True-colour wide-field confocal scanning is a novel technology, and we are confident that EyeArt performance can be further optimized for this exciting imaging innovation.”