UAMS Leaders Spotlight AI Innovation at TRI AI Research Symposium
| UAMS College of Medicine Dean Steven A. Webber, M.D., joined other UAMS leaders and researchers April 29 for the Translational Research Institute (TRI) AI Research Symposium, highlighting the use of artificial intelligence in biomedical research.
“Many of our faculty are already leading the way in several areas of AI-driven research, and that’s clearly something we want to see at UAMS,” Webber said in his welcome address.
Steven A. Webber, MBChb, MRCP
The symposium was led by Fred Prior, Ph.D., distinguished professor and chair of the College of Medicine Department of Biomedical Informatics, and Mathias Brochhausen, Ph.D., professor and vice chair for Academic Programs and Faculty Development in the Department of Biomedical Informatics. Prior also leads TRI’s Comprehensive Informatics Resource Core and Brochhausen is TRI’s associate director for Strategic Collaborations.
“It’s a great opportunity to share with you some of the research that we are doing in this field,” Prior said. “We use AI tools every day; we develop new ones, and we’re helping UAMS learn how to use these technologies to advance clinical practice and biomedical research.”
Webber, also UAMS executive vice chancellor, said that AI is now touching almost every facet of health care, including research, diagnostics, risk prediction, clinical documentation, medical education and revenue cycle enhancement.
“We have to embrace these changes and make them work for us in a positive way,” he said.
The three-hour symposium drew over 50 attendees and featured presentations by:
- Yasir Rahmatallah, Ph.D., associate professor, Department of Biomedical Informatics
Parkinson’s Disease Identification by Voice Assessment Using Pre-trained CNN - Jonathan Bona, Ph.D., assistant professor, Department of Biomedical Informatics
AI and Natural Language Processing - Aaron Kemp, MBA (Ph.D. candidate), co-director, NeuroCognitive Dynamics Lab; instructor, Department of Biomedical Informatics
Using Artificial Neural Networks to Mine for Markers of Abnormal Neural Network Activity among People with Parkinson’s Disease
Webber praised the research integration involving biomedical Informatics, clinical informatics and the clinical and translational science infrastructure provided by TRI.
Wendy Ward, Ph.D.
“We have to continue to further strengthen that integration so that the AI research that’s going on here starts to directly impact the health of our patients that we serve, and I know we are poised to do that,” he said.
Wendy Ward, Ph.D., professor and associate provost for faculty with UAMS Academic Affairs, described how AI is benefiting UAMS across its missions.
Examples include clinical tools that can draft Epic inbox responses, recommend assessments and diagnoses, generate discharge summaries and optimize billing codes. On the education front, she said UAMS has trained 100 educators in interactive AI modules, launched a journal club and student AI club and is developing a generative-AI course.
Both Prior and Ward touted the Department of Biomedical Informatics’ Creative Health AI (CHAI) Salon, an incubator of open-session workshops to develop AI solutions and collaborations. It has already produced two grant proposals and a Nature-Scientific Reports publication. A continuing medical education course on AI fundamentals is also in the works.
Fred Prior, Ph.D.
The symposium also included Prior’s call to action for cleaning, standardizing and curating research data to prevent “garbage in, garbage out.”
Prior warned that “garbage in, garbage out applies even more forcefully in the world of AI,” noting that algorithms require vast quantities of high-quality, well-labeled data to train effectively. He urged adoption of standardized data dictionaries and curated repositories to ensure consistent, reliable inputs.
Brochhausen added that machine learning is even more powerful when combined with ontologies.
“Ontologies organize and translate information and provide an enterprise-wide vocabulary, very much like a data dictionary, but with the added power to manage that dictionary, prevent contradictions, and support integration, reasoning and explainable AI,” he said.
Near the symposium’s conclusion, Prior announced that UAMS is building the legal and technical framework to offer researchers the use of large-language AI models that ensure patient data never leaves UAMS firewalls.
“If you give your data to a large language model on the Internet, it will happily reuse that data,” he said. “If it’s patient data, you’re spreading it all over the planet. Not a good thing.”
He hopes to roll out an upgraded model capable of handling images and text by summer, with hands-on training available in the fall.
TRI Director Laura James, M.D., praised Prior and Brochhausen for leading the symposium and for making it accessible to a broad audience.
“I’m excited about where UAMS is with artificial intelligence and the things that will be developed in the near future,” she said.