In Nancy Padilla-Coreano’s lab in the Evelyn F. and William L. McKnight Brain Institute, use of artificial neural networks, a form of artificial intelligence, is having a revolutionary effect on researchers’ ability to quantitatively measure behavior.
Padilla-Coreano, PhD, an assistant professor in the department of neuroscience, studies social competence, or how the brain helps individuals navigate their social world and interact with others. By studying social animals such as mice, she and her team hope to gain insight into human behaviors. One issue facing the researchers, however, is how time-consuming it can be. Until recently, social interactions in a mouse model were measured by hand, with trained researchers sitting for hours to watch videos and laboriously document behavior.
“Having people do the watching is super costly and very slow and inefficient, and people are not capable of quantifying to the level that computers can,” Padilla-Coreano says. “AI has accelerated our ability to quantify social interactions. If you cannot quantify something, you cannot study it.”
At the UF College of Medicine, AI is opening new doors for researchers like Padilla-Coreano. Across a wide spectrum of disorders, UF neuroscience and neuromedicine experts are implementing various AI techniques to enhance their research and clinical practice.
Learn about some of the neuromedicine research currently underway that merges minds and machines at the UF College of Medicine.
Addressing autism spectrum disorder
UF neuroscientists are studying autism in new ways, from Padilla-Coreano’s research to that of pediatric neurologist Brandon Zielinski, MD, PhD, who uses MRI to analyze the structure and function of brain networks as they develop. Zielinski investigates whether there is a developmental trajectory difference in specific brain networks in neurotypical children versus those with autism.
“I have many lifetimes’ worth of data, so I have to figure out how to process that data faster,” says Zielinski, an associate professor and chief of pediatric neurology who was recruited to UF last year as part of the Artificial Intelligence Initiative and will use UF’s HiPerGator supercomputer in his work. “That’s where AI comes in.”
For his latest project, which will follow adults with autism as they age, Zielinski will build AI frameworks to process data collected by his collaborators at the primary site, the University of Wisconsin-Madison. The study, part of the National Institutes of Health’s Autism Centers of Excellence program, will use brain imaging, blood tests and neuropsychological testing over time in a large group of adults, with and without autism, to examine changes in well-being, physical and mental health and brain structure and function.
Zielinski will also build the AI framework for the Autism Phenome Project, a large collaborative study headed by the University of California, Davis MIND Institute to identify and describe different types of autism.
Making an impact on movement disorders
AI applications for improving treatment for patients with movement disorders includes work by assistant professor Coralie de Hemptinne, PhD, MS, and biomedical scientist Jackson Cagle, PhD, researchers at the Norman Fixel Institute for Neurological Diseases at UF Health who have developed an algorithm to optimize deep brain stimulation, or DBS, a treatment that involves placing a thin wire in the brain in areas that control movement. Their technology, which received UF Innovate’s 2022 Invention of the Year award, predicts the best stimulation settings based on individual brain activity, shortening the wait to see improvement in symptoms.
Meanwhile, Joshua Wong, MD, an assistant professor in the department of neurology, is part of a team of interdisciplinary experts at the Fixel Institute specializing in DBS. The UF team has treated more than 1,500 patients using DBS to ease tremors, stiffness and slowness resulting from Parkinson’s disease, essential tremor or dystonia.
“Essentially you’re operating a minicomputer that has all these ways to deliver electrical energy to the brain,” Wong says. “When you calculate the possible permutations of the different options, it’s over a hundred thousand combinations.”
Enter AI: Wong is building a tool that will incorporate MRI scans and brainwave measurements along with brain recordings taken during DBS surgery, and the tool will calculate optimal settings from the outset.
Fixel Institute Executive Director Michael Okun, MD , along with colleagues from across the UF campus such as David Vaillancourt, PhD, a professor and chair in the UF College of Health & Human Performance, and Angelos Barmpoutis, PhD, an associate professor at the UF College of the Arts, are also using MRI and AI to advance treatment of movement disorders. They are testing a new AI tool to determine the precise diagnosis for patients who may have early Parkinson’s disease or one of two related but distinct Parkinson’s-like syndromes. Their tool will combine MRI images from a database of 315 patients at 21 sites across North America with a novel noninvasive biomarker technique that measures how water molecules diffuse in the brain and helps identify where neurodegeneration is occurring.
Moving the needle for memory disorders
Professor Jiang Bian, PhD, and assistant professor Jie Xu, PhD, of the department of health outcomes and biomedical informatics, are developing an AI tool to predict who will get Alzheimer’s disease up to five years before a diagnosis is made. Bian and Xu have published a study showing it’s feasible to use patient medical records containing known risk factors such as obesity, hypertension and high cholesterol to screen for future development of the disease, though Bian says more testing is needed before this tool is used by doctors and patients.
For Abbas Babajani-Feremi, PhD, the combination of AI and the Fixel Institute’s new magnetoencephalography, or MEG, scanner, represents a great opportunity: the potential to unlock information that could provide another way to predict who will get Alzheimer’s disease.
Babajani-Feremi, an associate professor of neurology and the director of the new MEG lab, is integrating machine learning with scans from the MEG scanner and from functional MRI, structural MRI and diffusion MRI in hopes of uncovering new biomarkers, or measurable indicators, of Alzheimer’s in the early stages. The MEG scanner, a noninvasive device encircling the participant’s head, is yielding valuable information about language, motor, auditory and visual processing, as well as where and when activity occurs in the brain.
“AI can highlight features you cannot see,” he says. “It can automatically search and find commonalities — for example, the volume of a specific brain region.”
Earlier detection of Alzheimer’s could guide treatment decisions and provide an opportunity to make changes in diet, exercise and sleep to potentially alter known risk factors for the disease, he says.
“For now, we don’t have a cure for Alzheimer’s disease, but if I knew that I might develop the disease in 10 years, I could take steps to mitigate its impact and prepare myself,” he says. “It’s very important we know as early as possible.”