Roche/Genentech Study Shows First-Time Proof That AI Can Detect Severity of DME

Source: Genentech

A new study published in the journal Investigative Ophthalmology & Visual Science suggest that artificial intelligence (AI) could be used to provide widespread, cost-effective eye screenings via telemedicine to assist ophthalmologists in improving vision outcomes for millions of people with diabetes who may not be getting regular eye exams, according to a statement from Genentech.

This is the first manuscript published as part of Roche/Genentech’s “Ophthalmology Personalized Healthcare” initiative that aims to combine meaningful large-scale data and AI technology to predict and prevent ocular conditions and preserve vision.

The study, led by Roche and Genentech scientists, demonstrated for the very first time, AI, specifically deep learning technology, can detect swelling in the macula and the severity of the swelling in individuals with diabetes, according to Genentech.

According to the American Academy of Ophthalmology, having a dilated eye exam yearly or as recommended by an ophthalmologist can prevent 95 percent of diabetes-related vision loss. However, all too often people with diabetes don’t get eye exams or vision screenings. And by virtue of the technology used, the severity of DME may not be reported. As a result, many people with DME may be under-diagnosed or treated late and are at risk for irreversible vision loss or blindness. 

Currently, the best way to diagnose DME is through the use of optical coherence tomography (OCT), which takes 3-dimensional, cross-sectional images of the macula. However, due to its cost and technical requirements, it is often not used in screening programs or telemedicine – which currently utilize two-dimensional color photos called color fundus photos (CFPs). The 2-dimensional nature of a CFP images can make detecting the severity of DME difficult. In order to address this limitation, the Roche/Genentech research assessed how deep learning can automatically view CFPs to accurately detect DME and determine its severity.

In the Roche/Genentech study, scientists used nearly 18,000 CFPs and their associated OCT images captured during Genentech’s past phase 3 DME studies to develop and assess the performance of deep learning algorithms. Results of the study showed that the best deep learning algorithm was up to 97 percent accurate in detecting DME severity using CFP images alone. Such results underscore the promising potential of AI in increasing screening capacity via telemedicine with appropriate triage to assist ophthalmologists in improving vision outcomes for a large population of patients who may not be getting comprehensive eye exams.   

The study adds to the growing literature about the use of AI in ophthalmology. It also sheds light on how Roche/Genentech can utilize its vast clinical trial database to develop AI algorithms to predict the presence of disease, risk of disease progression, and response to treatment; all of which could be supplied to ophthalmologists to deliver higher quality personalized healthcare.

Related Content