University of Tennessee faculty use artificial intelligence to solve problems in and outside the classroom

At the University of Tennessee, Knoxville, researchers are harnessing the power of artificial intelligence to deliver practical solutions to real-world challenges—from enabling more precise surgeries and quicker scanning for cancer to preparing students and workers for success in an increasingly AI-driven future.

UT's AI Tennessee Initiative brings together academic, industry, and community partners to leverage the benefits of AI across all disciplines and economic sectors and to make life and lives better across Tennessee through AI research and education.

"We have seen an explosion in demand for AI across every industry," said Vasileios Maroulas, associate vice chancellor and director of AI Tennessee. "When ChatGPT launched in 2022, the public perception of AI shifted overnight from something reserved for data scientists and engineers to a tool anyone could explore. That moment unlocked a wave of innovation, collaboration, and urgency."

As AI has become a go-to tool in settings from the production floor to the executive suite, AI Tennessee is working with industry partners to identify real-world problems that UT's AI researchers are helping to solve.

"By harnessing the expertise across our institution—and through close collaboration with partners across Tennessee—we are focused on AI that is actionable, deployable, and built to make a difference," Maroulas said. "AI Tennessee is expanding this impact statewide by aligning research, education, and workforce development efforts to meet the evolving needs of our communities and industries."

Giving students a better AI option

"No one can stay immune to the AI revolution," said ChuanRen (Charles) Liu, an associate professor of business analytics and statistics in UT's Haslam College of Business.

In March, Liu launched his own chatbot to help students find the information they need in the online textbook he and his colleagues authored for a class in data mining and business analytics. Students can click a "Chat with Me" button and talk to the textbook just as if they were talking to a professor.

"With this AI chatbot, they can ask any question in a conversational way, and the chatbot will only give them answers that are based on the textbook, not from anywhere else on the internet," Liu said, noting that students who turn to outside AI sources could get wrong answers because "not everything on the internet is correct."

Before the introduction of Liu's chatbot, students had to email their professors or come to office hours to discuss their questions. The chatbot can improve faculty members' productivity and enhance student learning.

Liu has already introduced the chatbot to other business faculty members to encourage the adoption of such tools more widely. After more testing, Liu plans to share the tool with other academic departments on campus.

"I'm happy to help because it's so useful," Liu said. "We can use AI to improve the learning experience for our students and also help ourselves get more work done."

When he's not teaching, Liu is leading several projects involving AI and business analytics. He is partnering with technology companies like HP to integrate predictive machine learning algorithms that reduce operational costs and increase profitability.

Liu also is working with colleagues in the Haslam College of Business and the College of Emerging and Collaborative Studies on several projects focusing on the impact of AI on jobs and workforce development. With the support of AI Tennessee, his research aims to determine future job requirements when businesses develop and adopt AI technologies.

"AI is developing so fast, and it's impacting all industry sectors, all businesses, and all organizations," Liu said. "We need to help workers be more open-minded about it because soon AI will be a universal tool for many professionals. No one puts 'I can use a calculator' on their resume. Being able to use AI is becoming a fundamental skill like we use calculators."

Training medical students with mathematical precision

A mathematician in Knoxville and a plastic surgeon in Memphis have teamed up to improve the outcomes of cleft lip surgeries on infants.

Cleft lip and cleft palate are birth defects that occur when a baby's lip or mouth don't form properly. In the United States, about 1 in 1,050 babies is born with cleft lip, with or without cleft palate. Surgical repair can help restore function to the lips and mouth and may improve the child's breathing, hearing, and speech and language development.

Theodora Bourni, an associate professor of mathematics in UT's College of Arts and Sciences, is applying her expertise in differential geometry with AI to help physicians and medical students visualize and practice cleft lip surgery.

"Geometric flows are differential equations that allow the user to see how geometric objects can take shape," she explained. "I am working with Dr. Robin Evans, an assistant professor in the UT Health Science Center College of Medicine in Memphis who specializes in cleft lip surgery, to standardize the procedure."

In repairing cleft lip, a surgeon first makes small dots called anatomical landmarks, indicating where specific cuts will be made. The surgeon then detaches and rearranges the tissues, using stitches to bring the left and right sides of the cleft together.

"The more experience a doctor has, the more symmetrical the outcomes," Bourni said. "In this project, we are using AI with ideas from differential geometry to predict the landmarks and eventually the outcome of the surgery."

Bourni received a grant from AI Tennessee in 2024 that allowed her and her Ph.D. students to train an initial algorithm for landmark prediction. Data from UT Medical Center in Knoxville will allow her to improve the prediction algorithms and train medical students to perform more symmetrical and successful surgeries.

Bourni's project could serve as a model for how AI learning collaborations can transform not just one medical procedure but entire domains of industry practice.

Detecting disease for earlier interventions

"I don't like to do science just for the sake of science," said Hector Santos-Villalobos, an assistant professor of computer science in UT's Tickle College of Engineering. "My goal as a researcher is to solve stakeholder problems."

That drive to create new solutions led Santos-Villalobos from successful stints at Oak Ridge National Laboratory and Amazon Prime Video and Studios to UT, where he is using AI and data science to tackle projects ranging from earlier Alzheimer's detection to faster cancer screenings.

"I use AI to expedite research processes from ideation to implementation, and I teach how to build AI systems to answer complex questions," Santos-Villalobos said.

For example, doctors use a variety of tools to diagnose patients with Alzheimer's—PET scans, MRIs, CT scans, bloodwork, and cognitive tests—but those sources of information on a single patient can come in at different times.

"Alzheimer's patients' clinical data is vast and diverse, making it challenging for clinicians to find patient-specific correlations and causations," said Santos-Villalobos, who is partnering with Roberto Fernandez-Romero, director of the Pat Summit Clinic at UT Medical Center. "It takes AI to make these deep data connections such as patterns between PET scans, encephalograms, and cognitive tests."

Untangling complex data patterns with AI allows physicians more time to focus on the science of treating or curing Alzheimer's.

"The earlier we can detect Alzheimer's, the better chance patients have of enjoying a longer quality of life," Santos-Villalobos said.

Another of Santos-Villalobos's projects is a collaboration with Mohammed I. Quraishi, an assistant professor of radiology in the University of Tennessee College of Medicine in Knoxville, to use AI to screen for lung cancer.

"For radiologists, lung cancer screening is a workflow bottleneck as it requires substantial time and focus to find nodules as small as a few millimeters on the background of a large amount of normal lung tissue," Quraishi said. "This is not intellectually hard, but it is tedious and requires focus—so it's the perfect job for AI. It would significantly cut down on the time it takes to read the CT exam."

Santos-Villalobos created different state-of-the-art algorithms to detect anomalies in the lungs more quickly. His goal is to use the same approach to screen for cancer in other areas of the body, such as breast and mouth cancer.

"This goes beyond finding the nodules," Quraishi added. "AI can also provide more insights such as trending the size growth of the nodules across multiple exams or finding features of a lesion not readily seen by the human eye. All in all, this allows radiologists to concentrate on interpretation rather than the search process."

Provided by University of Tennessee at Knoxville