Parkinson's patients and their doctors usually can't know how the disease will progress until it already has. A UM study found AI could see it coming years ahead of time.
New research from the University of Miami suggests AI may help identify Parkinson's patients at increased risk for rapid decline. It can flag that years before those changes become apparent.
What the Study Actually Found
Published in npj Parkinson's Disease, the study found machine-learning models could predict which patients faced significant decline. That covers cognitive or motor worsening over the following three to five years.
"One of the most difficult things about Parkinson's disease is not knowing how it will go," said Dr. Ihtsham ul Haq, professor of neurology at the Miller School of Medicine and the study's senior author. "We wanted to use AI's ability to analyze many kinds of information at once to see if it could predict who would have a more rapid motor or cognitive decline."
The most striking finding wasn't about imaging at all. Some of the most valuable predictive signals came from clinical measurements neurologists already collect during routine visits.
An Interdisciplinary Build From the Ground Up
The project brought together UM neurologists, radiologists and computer scientists. Yelena Yesha, the Knight Foundation Endowed Chair of Data Science and AI, joined after Dr. Haq and co-author Dr. Tatjana Rundek approached her about applying AI to neurodegenerative disease.
"The whole premise here is to use AI to attack dementia," Yesha said. "We are using neurology, neuroimaging, radiology and many other areas of clinical expertise to attack a very difficult problem."
Research assistant professor Yusen Wu worked with Yesha to develop and validate the machine-learning models. They trained the models on MRI scans combined with clinical and symptom data.
"As an interdisciplinary researcher working across neurology and computer science, I've seen what AI can do when grounded in real clinical data," Wu said. "It finds patterns in disease progression that no single field would catch alone."
Why the Dataset Itself Mattered
Yesha credited the underlying clinical data as much as the modeling approach. "Using this amazing dataset that the University of Miami has based on their clinical experience with patients, we were able to come up with these exciting results that are AI-driven," she said.
Leonidas Bachas is dean of UM's College of Arts and Sciences. He framed the collaboration as a model for how progress happens in this field. "These types of interdisciplinary partnerships have tremendous potential to improve the lives of patients and their caregivers."
What This Means for Miami's Medical Research
A prediction tool doesn't cure Parkinson's. Wu was careful to frame it as a complement rather than a replacement for clinical judgment. "It doesn't replace clinical judgment. It sharpens it," he said.
For patients and families, that distinction matters less than the outcome. Getting ahead of a diagnosis years earlier, using data doctors are already collecting, is a real, usable AI application. It changes treatment decisions now, not someday.
Florida Sets AI School Rules. Miami Dade Already Acted