...despite the outstanding treatment outcomes in the controlled clinical trial setting, the medical community has struggled to sustain these results in the real world due to a need for efficient early diagnosis followed by frequent exams and treatment injections. 

Diabetes remains the principal cause of blindness in developed countries, and the frequency of vision loss due to diabetic retinopathy (DR) is on the rise and expected to reach epidemic proportions globally in the next few decades. DR is the most common microvascular complication of diabetes, and the condition can progress symptom-free in patients to a very advanced state.

Diagnosis of DR is based on an assessment of a 2D image of the retina, called a color fundus photograph, by a retina specialist. Today we are learning that machines can be  far superior to the human eye in evaluating retinal images, not only in speed but also in its’ ability to detect signals and patterns invisible to human eye. Pioneering work in that field has been done by Moorfield Hospital (London, UK), in collaboration with Google, by us (Arcadu et al., IOVS 2019) and by others.

In the work published this week, we used 2 Phase III clinical trial datasets from sham treated patients (RIDE and RISE) to predict the future fate of DR patients across a range of DR severity at presentation from single, initial 2D retinal images.

Once validated on other datasets, the key application of our pilot algorithm in real world is in its’ potential to effectively triage diabetes patients across all stages of disease to pinpoint those individuals who need most urgent specialist attention to diagnose and consider close follow up or early intervention if warranted. This is particularly important considering the rising global prevalence of diabetic eye disease and its projected impact on healthcare systems and society.