By Michelle Jaffee
Using artificial intelligence, University of Florida scientists have demonstrated a potential new way to improve medication management in Parkinson’s disease and other movement disorders.
Reporting in the journal Proceedings of Machine Learning Research, researchers from the McKnight Brain Institute, Norman Fixel Institute and UF Intelligent Clinical Care Center used AI to predict Parkinson’s medication needs up to two years in advance, along with confidence levels for the predictions.
“Currently, doctors rely on trial and error when adjusting Parkinson’s disease medications, which can lead to inadequate symptom control for serious side effects like involuntary movements derived from dopamine-replacement therapy,” said lead author Ricardo Diaz-Rincon, a doctoral student in the UF Department of Neuroscience.
Such involuntary movements may include facial grimacing, writhing and sudden twisting movements.
“Our method could tell providers not just ‘this patient will need X amount of medication’ but also ‘we are 80% confident in this prediction,’” Diaz-Rincon said. “This approach has the potential to transform how neurologists make treatment decisions, moving from guesswork to evidence-based care that could significantly improve quality of life.”
The research team, which also included senior author Benjamin Shickel, Ph.D., Adolfo Ramirez-Zamora, M.D., and Muxuan Liang, Ph.D., developed a machine-learning framework using deidentified electronic health records from 631 inpatients at UF Health from 2011 to 2021.
Their advanced statistical methods could lead to more precise and personalized treatment approaches, Diaz-Rincon said. Next, the team plans to incorporate outpatient data and additional clinical measurements, such as motor-function scores and quality-of-life assessments, among other ways to refine prediction accuracy.