Medicine’s new frontier

ith today’s patients continuously tuned into smart devices, it makes sense that soon, a medical chatbot you interact with online will be able to use conversational language to communicate with and educate patients in much the same way we now interact with Apple’s Siri and Amazon’s Alexa.

What may have sounded like science fiction just a few years ago is on the precipice of reality. UF’s strides in the field of artificial intelligence, including a $70 million partnership with the Silicon Valley-based technology company NVIDIA, are leading to research and industry innovations across all disciplines.

At the College of Medicine, experts say a focus on AI will result in better patient care, advancements in clinical research and new techniques to train the next generation of clinicians and medical scientists.

One way this has already happened is through the development of SynGatorTron™, an AI natural language processing model that generates synthetic patient data untraceable to real patients, which can be used to train the next generation of medical AI systems to understand conversational language and medical terminology.

“In the next decade of medicine, technology is going to be our partner,” says Azra Bihorac, MD, MS, senior associate dean for research affairs at the College of Medicine. “We need physicians who will dream up new solutions and work to understand how technology can deliver them in partnership with other specialists like engineers, data scientists and biostatisticians.”

Health care teams routinely collect significant amounts of patient information and data. While that information can be useful for patient outcomes and research, Bihorac says, sifting through millions of data points to draw meaningful conclusions is time-consuming and nearly impossible for already busy health care teams to complete on their own.

“We can give physicians from all backgrounds the same training and tools,” says Bihorac, also the R. Glenn Davis Professor of Medicine, Surgery and Anethsiology and co-director of UF’s Intelligent Critical Care Center. “Then it really doesn’t matter if you’re in a small community hospital. AI can level the playing field for patients, no matter where they are.”

Bihorac and colleagues across the UF campus are involved in research that addresses how AI algorithms can help health care teams determine the potential risks for complications following surgery, based on routinely collected patient data and past procedures.

“The algorithms will help clinicians address patients’ needs rather than solely relying on their intuition and experience to determine what will be best for the patient,” says Tyler Loftus, MD, an assistant professor in the department of surgery and an acute care surgeon at UF Health Shands Hospital.

The algorithm developed by UF researchers, called MySurgeryRisk, has been internally validated at hospitals in Gainesville and Jacksonville and will next be tested at several health care centers in the Southeast and the Midwest.

The researchers are ensuring that potential biases are addressed by testing the algorithm in areas with varied patient population demographics.

“Mitigating bias in care is an important step in this process,” says Loftus, who completed his residency and fellowship training at UF. “For example, the data that come out of Gainesville will look different from the data we gather in Jacksonville at an urban hospital in the heart of a big city.”

Through a partnership with NVIDIA, faculty at the College of Medicine are leading the nation in developing novel ways to incorporate informatics and data science into health systems.

One application is having AI algorithms synthesize vast amounts of patient data to identify potential participants for clinical trials, says Jiang Bian, PhD, a professor in the department of health outcomes and biomedical informatics, director of the college’s Cancer Informatics & eHealth Core program and chief data scientist for UF Health.

“It is often difficult to find enough participants, and AI can help clinicians proactively identify patients who would qualify for trials and do more targeted recruitment,” Bian says. “This can help speed up the process of recruitment and lead to more reliable data with qualified participants.”

Routinely collected patient data can also be used by researchers who are trying to better understand social determinants of health. For example, AI data synthesis can help researchers more easily study social risks that are linked to the development of Alzheimer’s disease, such as marital status and education level.

“AI can comb through patient data and find people who might be at a greater risk of developing Alzheimer’s, for example, and they can be monitored to see how their health progresses and may also become eligible to participate in clinical trials,” says Yonghui Wu, PhD, director of natural language processing at the UF Clinical and Translational Science Institute.

Reducing error
Understanding the limitations of AI-generated conclusions and ensuring that implicit biases are being addressed is at the forefront for UF researchers. AI systems developed at UF routinely undergo rounds of study using different populations to better understand how algorithms might change across populations. Researchers also review data sets to flag and potentially remove pieces of information for study that have minimal impact on causality.

Increasing doctor-patient engagement
Medical chatbots like SynGatorTron™, developed by UF researchers, can use conversational language to communicate with and educate patients in much the same way people now interact with Apple’s Siri and Amazon’s Alexa. SynGatorTron™ can generate synthetic patient data untraceable to real patients, which can be used to train the next generation of medical AI systems to understand conversational language and medical terminology. Having a medical chatbot available in a health care system can decrease barriers to care for patients after hours and allow for real-time conversations about their health, no appointment necessary.

Detecting and diagnosing disease
Using a branch of AI known as machine learning, UF researchers have zeroed in on the specific bacteria suspected of causing depression and high blood pressure. In another UF study, researchers combined an AI computer algorithm with MRI brain scans to accurately predict whether people with a specific type of early memory loss will go on to develop Alzheimer’s disease or other forms of dementia.

Personalizing treatment and informing care
UF researchers have developed and successfully tested an AI system that delivers streamlined and timely details about crucial changes in a patient’s condition. The system, known as Deep Sequential Organ Failure Assessment, or DeepSOFA, works by collecting, organizing and presenting a patient’s medical data so doctors can make nimbler, better-informed decisions. DeepSOFA buys critical time by indicating which intensive care unit patients may need a lifesaving intervention to prevent potentially fatal conditions, and it can be a powerful predictive tool to help doctors determine how a patient’s condition is trending and what may be causing that change.

While the College of Medicine’s clinicians and researchers are pioneering the use of AI in the medical field, students and trainees are taking a deep dive into one of the nation’s first curriculums centered on the intersection of AI and medical education.

“AI is becoming part of our daily lives in so many ways that are sometimes obvious, and sometimes less readily apparent,” says Patrick Tighe, MD ’05, MS, the associate dean for AI application & innovation and an associate professor of anesthesiology at UF. “It’s important for physicians to understand how these tools work from both a theoretical perspective and a more pragmatic perspective, to understand how AI can be applied in clinical practice as well as for broader population health.”

With these goals in mind, Tighe is one of the leaders on a team of College of Medicine faculty who are developing a comprehensive AI curriculum aimed at clinicians, students and trainees who want to learn more about ways to incorporate data and machine learning into patient care. Chris Giordano, MD, an associate professor of anesthesiology, and Francois Modave, PhD, a professor of AI in the department of anesthesiology, are also leading the development of the curriculum.

The first online course, which was created in partnership with the UF College of Education, launched this spring for various members of the college community and will continue to roll out into 2023. It allows those with a clinical background to learn the basics of AI and how to become further involved in AI research.

Additionally, PhD candidates in the Graduate Program for Biomedical Sciences can now collaborate with some of the university’s preeminent AI researchers as part of the Emerging Research Scholars-AI PhD Program. The program, which offers additional support in AI from the Office of Research for students’ first two years of study, connects students with AI mentors and teaches the fundamentals of AI development and implementation.

However, Tighe notes, AI tools are limited by the information available to the machine. The experience trainees and early-career faculty can gain from working with senior faculty is just as valuable in the process of training others about how to apply AI in medicine.

“There’s a lot of wisdom, especially that our senior physicians accumulate, that is highly valuable and can be complemented by AI tools,” he says. “And when we talk about excellence in clinical practice, we’re talking about more than just the cognitive tasks. It’s the empathetic nature, the humanism, the compassion we share for another human being as a physician. There’s nothing artificial about any of that.”