John A. Moran Eye Center researcher , is part of a big leap forward for the use of artificial intelligence in eye care.
Cofounded by Dubis, German-based health tech company has gained for an AI algorithm that assists ophthalmologists treating so-called “wet” or neovascular age-related macular degeneration (nAMD).
The company’s Treatment Planning Support (deepeye® TPS) product is the first predictive AI tool for ophthalmology therapy management approved for clinical use in the Western world.
“This is an exciting milestone in the collective global effort to ethically source and provide physicians with data in a way that can advise treatment and improve care,” says Dubis, who is applying to have deepeye TPS tested at Moran Eye Center clinics later this year as a step toward FDA approval.
Working to Personalize AMD Treatments
AMD is a leading cause of blindness among people age 55 and older, and patients with the neovascular form experience abnormal growth of blood vessels in the eye that can leak and rupture. Specialists treat neovascular or “wet” AMD with injections of medication aimed to prevent this growth and use retinal imaging to evaluate progress.
But not all patients respond to the same treatment regimen equally; treatment must be personalized. If appointments are scheduled too far apart, or if patients feel they are coming in too frequently and miss some appointments or even drop out of treatment, they could lose vision.
To create deepeye® TPS, Dubis and colleagues used tens of thousands of 3D scans of the retina combined with medical records to analyze the progression of nAMD in correlation with treatment regimens. The resulting algorithm, he says, helps physicians make two decisions, acting like a second expert reader of the scans.
- Does the person need to receive the next treatment sooner than the last one, or can the time between appointments be extended?
- How many treatments will the patient need over the next 12 months?
“The idea is that it will allow patients to keep their vision as long as possible and ensure they are only spending time going to a clinic when they need to,” says Dubis. “For physicians, it optimizes clinic flow, provides reassurance for the treatment course, and possibly helps in patient education or justifying therapy switching.”
To gain approval in Europe, deepeye® TPS was tested with an international group of over 300 patients (more than 2,000 visits). The physicians decided treatment plans independently based on scans, then consulted with deepeye® TPS after the fact to determine if their decisions aligned and if not, why.
When ophthalmologists participating in the study at Ludwig Maximilian University Eye Clinic in Munich recommended a treatment different from deepeye® TPS, they agreed with the recommended changes 56% of the time.
Moran Eye Center retina specialist and researcher Eileen Hwang, MD, PhD, is a highly trained vitreoretinal surgeon who regularly treats wet AMD patients. She says there is “definite potential” for improved accuracy as ophthalmologists must experiment to identify the optimal time interval between injections for each patient.
“Most physicians treating wet AMD are trying to personalize medicine by testing and seeing what works, but if we have some predictors, that could help us get there more quickly or more safely,” she says.
Hwang cautioned against viewing AI as a panacea.
“Nothing is going to be 100 percent,” she says. “We need to keep our minds open that even AI is a prediction that could be wrong, just like when a physician makes a decision to inject or not to inject a patient on a given day.”
Additional AI Initiatives
The new deepeye® TPS is one of several AI-related projects for Dubis, who joined the Moran Eye Center in 2024 as a renowned expert in image and health data analysis who broadly consults across the image analysis and biopharmaceutical space. Dubis works with several international bodies on the evolution of policy and health technology regulation.
At the Moran Eye Center, has created the Moran Phenotyping, Imaging and Advanced Technologies (PHIAT) database. This anonymized AI database can be queried by researchers to study myriad questions related to ophthalmology. For example, a researcher might query the database to predict which patients with diabetes are most at risk of getting diabetic eye disease, or to create an algorithm that helps doctors identify when one drug isn’t working and it’s time to try another.
Dubis emphasizes the gathering of ethically sourced data. In PHIAT, individual health records are de-identified, and the confidentiality of the data is strictly protected. Dubis explains that PHIAT is unique among academic medical centers.
“While other universities have limited databases open to research access, the data is preconfigured, whereas PHIAT is not,” says Dubis. “This allows for a much broader range of inquiry and investigation.”
Moran Eye Center CEO and Distinguished Professor and Chair of Ophthalmology Randall J Olson, MD, says the institution is excited to push the field forward in the areas of data and AI. Huntsman Cancer Center is already home to the Utah Population Database, one of the world’s richest sources of information on more than 11 million individuals that supports research on genetics, epidemiology, demography, and public health.
"Ophthalmology in particular is a field where we rely heavily on imaging of the eye, which has seen tremendous advances in the past decade,” explains Olson. “AI can assist with correlating these images to disease progression and to find those patterns that will better inform our treatment plans. Simply put, this is the future, and we are investing in it.”
