When Artificial Intelligence Meets Health

AI generated.

As artificial intelligence shakes up the world, the University at Buffalo—and the School of Public Health and Health Professions—are gearing up to harness and advance this powerful technology for the greater good. 

The effort is benefiting from a major New York State initiative that places UB at the center of an innovative AI research program, which, ultimately, positions social good firmly as its baseline.

UB has a rich, 40-year history in AI research and is recognized as a leader in the field. Over 200 university researchers work on AI and data science projects. Those projects include work that most people would associate with AI like robotics, cybersecurity, etc. But researchers and faculty in the health sciences areas are applying their expertise to developing and using AI in areas such as drug discovery, diagnostics, women’s health and much more.

All of this will take a giant leap forward thanks to Empire AI, a groundbreaking consortium of New York’s leading educational institutions, including the State University of New York, that New York State Governor Kathy Hochul and the state legislature launched this year. Empire AI aims to harness the power of AI to address major societal challenges in fields like health care, education, social justice and climate change.

Significantly, Empire AI committed $275 million to create a state-of-the-art AI computing center at UB. The project has also secured over $125 million from SUNY and the other founding institutions Columbia University, Cornell University, New York University, Rensselaer Polytechnic Institute, the City University of New York and the Flatiron Institute. Funding is also coming from private partners, for a total investment of more than $400 million.

The governor emphasized the importance of the initiative, stating, "Whoever dominates the AI industry will dominate the next chapter of history."

AI and society

Empire AI's commitment to social good and health care is a central part of its mission, according to SPHHP Dean Jean WactawskiWende, PhD. She emphasizes that the consortium plans to leverage AI not just for economic growth but also to address serious societal challenges and improve people's health.

Said Venu Govindaraju, PhD, UB's vice president for research and economic development, “With UB as the home of Empire AI, our Institute for Artificial Intelligence and Data Science, and our six health sciences schools plus engineering, UB clearly has the talent and experience to lead AI in health….”

Indeed, Govindaraju’s office and the Office of the Vice President of Health Sciences have already put rubber to the road by offering funding for multidisciplinary projects at UB using artificial intelligence to enhance health care. Researchers will submit proposals to compete for the interdisciplinary seed funding for their projects.

In health, as in so many arenas, the potential applications of AI are numerous and promising. Among many others: 

  • Researchers could develop AI systems to assist in early disease detection, personalized treatment plans and drug safety. 
  • AI could also help in predicting patient outcomes or optimizing hospital operations. 
  • In the realm of community health, AI tools could analyze data on systemic health disparities, helping policymakers make more informed decisions to address these issues. 
  • In environmental health, AI models could be used to improve air and water monitoring, disease outbreak prediction due to climate change and more.

In SPHHP, AI is already playing a role in research and in education, with faculty and investigators implementing it in small and more substantial ways.

Underpinning AI

Landisttel ehadshot.

Statistics, for instance, works hand in glove with AI. 

“While we might consider some AI methods (e.g., driving of autonomous vehicles) as outside of the realm of statistics, many potential applications of AI methods require statistical models where the advancement of statistical methodology is critical to their success,” explained Doug Landsittel, PhD, chair of the Department of Biostatistics. He offered an example in the potential use of neural networks—the heart of AI—for clinical prognosis using large sets of data from electronic health records.

Landsittel and other researchers in the department are already working to refine the statistical methods that underlie AI. He notes the many AI challenges that need statistically valid methods to address issues like predicting the risk for undergoing certain medical procedures. He also noted that statisticians are more frequently working with other fields to add the statistics perspective.

“We really need multidisciplinary teams,” he emphasized. If a target of AI, for example, is the prognosis and predictions of treatment effectiveness, “this is a very multidisciplinary problem that needs a multidisciplinary approach. I’d like to see AI development run like we run a clinical trial, with statisticians, clinicians, data collectors, informaticians, engineers and computer scientists all involved from day one.”

Ultimately, Landsittel emphasized, statistical methods are essential for the advancement of AI, and the Department of Biostatistics is already waist deep in the effort: “Our faculty have published, and will continue to publish, and apply for funding in these topics to advance the utility of AI methods." 

Accelerating AI use

LaMonte headshot.

Research Professor Michael LaMonte, PhD, is making use of an AI-aided device in a project with the national Women’s Health Initiative. It involves wearable activity monitors—accelerometers— to measure daily activity patterns and intensities in women between 63 and 99 years old. 

“We train the accelerometer based on several criteria or ‘gold standard’ sources of truth to recognize patterns in the raw acceleration data that represents different types of movement such as changing posture from lying to sitting to standing; walking at different stride lengths and rates; and intensities of varying magnitude both in absolute and relative contexts,” LaMonte explained. 

“We also are using AI to process large amounts of electrocardiogram (ECG) data obtained from the same women while wearing a cardiac patch on their chest. In this way, we can measure heart rate variability, an indicator of autonomic nervous system regulation of heart rate and blood pressure. AI also allows us to align the beat-to-beat heart information with the second-to-second acceleration data from the accelerometer.”

To LaMonte’s knowledge, this is one of the first, if not the first, large epidemiological study on older women enrolled from the community setting to collect high-dimensional data in the cardiac and movement domain and apply AI as a tool to process and learn from the data in a manner that might not be possible in the absence of AI methodologies.

In fact, the WHI project is a great example of the kind of multidisciplinary efforts that Biostatistics’ Doug Landsittel advocates. 

“I am not expert in AI,” LaMonte said. “The work we are doing is based on strong and mutually fulfilling partnerships with AI scientists and bioinformaticians together with our epidemiological, biostatistical and clinical team members.”

Promoting AI ethics

Kruger headshot.

Though she sees definite possibilities for AI in public health, the forays of Clinical Associate Professor Jessica Kruger, PhD, MCHES, into AI are primarily in the realm of education. Director of SPHHP’s Teaching Innovation and Excellence and a member of UB’s Task Force on Generative AI in Teaching and Learning, she has a keen perspective on what’s happening from a teaching standpoint—in her words, “a lot.”

The UB AI task force made concrete recommendations: “Faculty should choose what they want to do, but they need to make explicit what they expect from students,” Kruger explained. For instance, she said that the syllabi for SPHHP’s Master of Public Health classes will soon have explicit language about student use of AI, mostly because students get differing messages from different sources.

Kruger believes that teaching UB students AI literacy— its ethical uses and how they can use it day to day—is vitally important. 

“There’s a learning curve for students and faculty. AI isn’t perfect, but it is a tool in the toolbox, whether it’s spell check or it helps you think about a broader topic,” she said.

“Our school is actually going to build a class on AI for students. By fall, we’ll have a short tutorial on what AI is and information on future possibilities. I could see a whole course on AI and public health.” That’s a topic she’s already written about in a paper: “We said that you cannot not use this. You need to understand its capabilities and become familiar with it. We also highlighted why it’s important to teach students how to use it.”

On a more concrete AI note, a student is building Kruger a chatbot for a course she teaches. 

“We’ll feed it the syllabus along with questions that students might be hesitant to ask about—can I get an extension on my project, would you provide me a recommendation, etc. This sort of use is a unique application of AI in teaching and supporting student success.”

Collaborating on a huge effort

Hageman Blair headshot.

The associate director for education of UB’s Institute for Artificial Intelligence and Data Science (IAD), Associate Professor of Biostatistics Rachael Hageman Blair, PhD, has been working with AI “for a long time.” Her expertise is in data mining, networks and clustering methodologies that makes data from differing sources easier to work with, and she has published several papers in these areas. 

Hageman Blair's research exemplifies the kind of collaborations that make effective use of AI. One study, funded by a National Science Foundation grant and piloted by SPHHP, is trying to figure out how a common byproduct of yeast used in medicines, cosmetics and many other products can be produced in larger quantities to keep up with demand. Study investigators are using AI to rapidly test millions of genetic mutations of yeast to determine which might best increase the byproduct’s production and should be tested by lab colleagues. 

The other focus for Hageman Blair is her role overseeing the educational directives of IAD, which she’s been part of since its inception. With three master’s and PhD programs, the institute enrolls 800 to 900 students every year in the 18-month program, which includes SPHHP-led biostatistics and statistical data mining courses. The institute also runs a summer series of short courses designed to meet the needs of students and industry and a signature annual conference, IAD Days.

Hageman Blair is also involved in other moves to grow UB’s footprint in AI. She recently was chair of a STEM group for the Office of the Vice President for Academic Affairs’ Task Force on Generative AI in the Classroom. 

“Every school and instructor needs to be aware of how both students and educators use AI,” she said. “Focus groups and surveys suggest that the perspective from students is positive, and that professors are more skeptical and have concerns over their ability to evaluate learning outcomes.” The task force just released a report to stakeholders on generative AI in the classroom and provided additional awareness on issues such as limitations in AI detection, and bias in tools such as plagiarism detectors. 

Ultimately, Hageman Blair feels the benefit of Empire AI is huge. "I don’t think we even realize it yet. The recognition for UB has been remarkable. We’ve done so much great work in the institute, and this will boost our capacity to do the massive simulations that require so much computing power. Other schools don’t have that edge.”