What if one could create a powerful tool, using artificial intelligence (AI), that would help clinicians more accurately diagnose, predict and treat the disease process in an individual patient’s retina? That is the question that T.Y. Alvin Liu, M.D., assistant professor of ophthalmology and retina specialist, has set out to answer as the director of the new Wilmer Precision Ophthalmology Center of Excellence.
“I like to take what I think of as the Elon Musk approach,” he says. “Instead of looking at the existing paradigm and infrastructure and asking what he can do with it, Musk [the renowned creator of Tesla and SpaceX] identifies his destination and works backward. So my question is: What kind of data, process and innovations do I need to get to my destination?”
Just before the pandemic started, Liu had a chance meeting with Caroline Popper, M.D., M.P.H., chief business officer of the Johns Hopkins Precision Medicine inHealth initiative, which is encouraging and offering support to physician-researchers throughout Johns Hopkins to create precision medicine centers of excellence to develop tools to answer specific questions in their various specialties.
“The idea behind precision medicine is that not everyone is the same,” says Liu. “So, we want to harness our exponentially expanding ability to process and analyze massive amounts of data in order to finetune treatment for individual patients.”
Liu says he was inspired by his meeting with Popper to begin the application process, and the Wilmer Precision Ophthalmology Center of Excellence was approved in March 2021. The center — which builds on earlier capabilities of the Wilmer Artificial Intelligence Research Network that Liu helped establish in 2019 — adheres to several principles he sees as fundamental to all major advances in medicine. “First, they must benefit our patients. Second, they should be scientifically interesting. And, finally, they should be commercially viable so they have the potential to be self-sustaining.”
AI is, simply put, the ability of a computer to mimic human intelligence. Machine learning and deep learning are subtypes of AI in which an algorithm is generated based on lots and lots of data, which the computer can then use to improve its predictions. Deep learning is exceptionally good at analyzing the complex data behind images and has the potential to make increasingly accurate predictions based on imaging data — which makes it well suited to ophthalmology because the field relies so heavily on imaging.
The advantage of tapping into Johns Hopkins Medicine resources is obvious, Liu says, because it allows Wilmer researchers “access to data in a secure, centralized place … and the data can be further transported to cloud-based Microsoft GPUs [graphic processing units] — special computer hardware that’s especially good for deep learning analysis.”
What will it mean for patients? Liu envisions patients coming into his office, getting a scan that’s read and analyzed instantly, and leaving with a personalized prediction that will guide treatment, including how quickly their disease might progress. The technology, he says, “will be better than humans, detecting things we can’t detect. But it’s not that your retina specialist is now a robot. It’s that I now have a very smart assistant, freeing me up to spend more time connecting with my patients and doing what humans do best.”
But first, Liu’s team has to create the AI infrastructure to do the kinds of mass data processing he envisions. “Just to give you a sense of the challenge, the most frequently used imaging modality in ophthalmology is optical coherence tomography, or OCT.” At academic medical centers like Wilmer, there are tens of thousands of OCT images produced every year, he notes. Typically, “researchers can usually only download one image at a time, which means images for a cohort of 300 patients can take weeks to download,” he says. “One thing we are about to complete is a new [virtual] pipeline with which we can download all the images for 3,000 patients in a day or two.”
The goal is that Liu’s project will inspire other researchers at Wilmer, in every division, to take advantage of and build on the tools his team is creating.
“We announced [the Center of Excellence] at a faculty meeting last fall, and I have already been approached by people, both in other divisions and within the Retina Division, who are interested in what these tools can do to support their research,” says Liu.
Johns Hopkins Precision Medicine Director Antony Rosen, M.B.Ch.B., who is also vice dean for research at the Johns Hopkins University School of Medicine, says he is delighted with the work being done at Wilmer. “Our evolving precision medicine approach is to build clusters of precision medicine centers of excellence in a specific area that use the same types of measurements, and then allow the tools developed in one center to be applied to other areas. Since so much done at Wilmer uses high-information content measurement of many different aspects of the eye, it makes sense to harness the efficiency of bringing those measurements onto the platform once, and well.”
Last year, Chief of the Oculoplastics Division Nicholas Mahoney, M.D., was named vice chair of information technology and data science at Wilmer, and he was tasked, he says, with “fostering our data science program and taking better advantage of the platform Johns Hopkins Precision Medicine has been putting together as a resource for everyone at Hopkins.”
Mahoney, who is working closely with Liu, says their data science goals at Wilmer are complex, ambitious — and very exciting. “By connecting people at Wilmer who are actively working on their own complex data-oriented projects with Dr. Rosen and his team, we can enable Wilmer researchers to build complex tools and access tremendously powerful data. We are working on getting everybody connected,” he says.
“We are living in a very unique time,” Liu says. “People often say AI is the fourth industrial revolution, and I’m a believer. It’s going to change medicine, and definitely ophthalmology, as we know it.”