By Michael Aubele
Photo credit: Mason Markland
A brief introduction into the basics of artificial intelligence (AI) began two days of discussions about AI’s use in research and clinical care in Pittsburgh, how it’s being applied in education at the University of Pittsburgh and how it could shape the future of medicine. The conversations unfolded Jan. 29 and 30 at the Digital Health Summit, an event intended to bring together digital medicine leaders in health care delivery and education.
Organizers said having two universities in the city well known for their medical and computational research and leadership in education in those fields were drivers in putting the summit together. “Pittsburgh genuinely is a living digital health laboratory,” said moderator Ken Nischal, a Pitt professor of ophthalmology, above. “There are so many digital platforms, so many institutions, so many startups coming from this region—we all interact and we all collaborate. It makes us so special.”
Thought leaders, researchers (including Pitt and Carnegie Mellon University students), private developers and legal experts discussed the growing influence of AI in health care. They covered topics such as machine learning in research, how digital twins could shape the future of personalized care and the potential for AI to replace humans.
Multiple physicians at the summit attested to the benefits of AI platforms like Abridge—a notable Pittsburgh startup and one of the summit’s sponsors—to help reduce physician burnout. Abridge built software to record and transcribe doctor-patient conversations, shaving administrative time and effort for physicians and “letting us focus on the more complex tasks we got into medicine to solve,” one attendee noted.
Hooman Rashidi, associate dean of AI in medicine at Pitt’s School of Medicine, has introduced a curriculum that introduces medical students to AI and enhances their ability to use it. The mission is to prepare future physicians for the changing technological landscape and to use the tools ethically. A goal in democratizing AI this way is to roll out similar curriculums across the health sciences, he said.
Pitt medical students already are using AI for varied pieces of research, including to determine how much patients trust it. Shalia Fye, a second-year student, is working with pregnant and postpartum patients to gauge their level of confidence in AI-generated written material.
“My particular focus and research interest is how large language models are increasingly used for patient educational materials,” she said. “These tools have real promise; they can simplify complex medical education, as well as improve health literacy. But the evidence on their effectiveness and acceptability is mixed. Even when AI-generated materials are high quality, patient trust remains a key barrier to adoption and is right now relatively understudied.”
Devante Kerr, another second-year student, presented on using machine learning to determine what pre-existing conditions could influence cognitive decline in patients with schizophrenia. Kerr and his team worked with data from the All of Us Research Program, a vast National Institutes of Health initiative, and applied two types of models to the question. While the results between the models differed, what overlap there was identified several diseases that could lead to cognitive decline in these patients. In addition to depressive disorder, Kerr’s team also identified pure hypercholesterolemia and hyperlipidemia as important predictors that may contribute.
Zach Spears, a student researcher, explained how Pitt researchers are using AI to create metamaterial rods used in spinal fusion surgery. He said current implants used in spinal fusion don’t replicate the natural motion of the spine and that a solution to the problem exists in using engineered mechanical metamaterials that allow for tuning of the mechanical properties so that they are extremely lightweight but strong.
Beyond the more than a dozen pieces of AI research presented at the summit, experts weighed in on philosophical questions many are asking about how AI and how issues such as governance should come into play. One hot-button topic was the potential for AI to become less of a tool and more than a replacement for humans in medical care—a discussion that the experts said will be ongoing.
“You know, the question that is asked in so many meetings is kind of a silly question,” said Zafar Chaudry, senior vice president, chief digital officer and chief AI and information officer, Seattle Children’s Hospital. “‘Will AI replace clinicians?’ The answer is easy. It's no. I’ve heard, ‘AI is not going to replace me, but I may be replaced by a physician who understands and implements AI properly.’ So that’s important to know.”