News Release

Research shows AI technology improves Parkinson’s diagnoses

Peer-Reviewed Publication

University of Florida

Existing research indicates that the accuracy of a Parkinson’s disease diagnosis hovers between 55% and 78% in the first five years of assessment. That’s partly because Parkinson’s sibling movement disorders share similarities, sometimes making a definitive diagnosis initially difficult.

Although Parkinson’s disease is a well-recognized illness, the term can refer to a variety of conditions, ranging from idiopathic Parkinson’s, the most common type, to other movement disorders like multiple system atrophy Parkinsonian variant and progressive supranuclear palsy. Each shares motor and nonmotor features, like changes in gait — but possess a distinct pathology and prognosis.

Roughly one in four patients, or even one in two patients, is misdiagnosed.

Now, researchers at the University of Florida and the UF Health Norman Fixel Institute for Neurological Diseases have developed a new kind of software that will help clinicians differentially diagnose Parkinson’s disease and related conditions, reducing diagnostic time and increasing precision beyond 96%. The study was published recently in JAMA Neurology and was funded by the National Institutes of Health.

“In many cases, MRI manufacturers don’t communicate with each other due to marketplace competition,” said David Vaillancourt, Ph.D., chair and a professor in the UF Department of Applied Physiology and Kinesiology. “They all have their own software and their own sequences. Here, we’ve developed novel software that works across all of them.”

Although there is no substitute for the human element of diagnosis, even the most experienced physicians who specialize in movement disorder diagnoses can benefit from a tool to increase diagnostic efficacy between different disorders, Vaillancourt said.

The software, Automated Imaging Differentiation for Parkinsonism, or AIDP, is an automated MRI processing and machine learning software that features a noninvasive biomarker technique. Using diffusion-weighted MRI, which measures how water molecules diffuse in the brain, the team can identify where neurodegeneration is occurring. Then, the machine learning algorithm, rigorously tested against in-person clinic diagnoses, analyzes the brain scan and provides the clinician with the results, indicating one of the different types of Parkinson’s.

The study was conducted across 21 sites, 19 of them in the United States and two in Canada.

“This is an instance where the innovation between technology and artificial intelligence has been proven to enhance diagnostic precision, allowing us the opportunity to further improve treatment for patients with Parkinson’s disease,” said Michael Okun, M.D., medical adviser to the Parkinson’s Foundation and director of the Norman Fixel Institute for Neurological Diseases at UF Health. “We look forward to seeing how this innovation can further impact the Parkinson’s community and advance our shared goal of better outcomes for all.”

The team’s next step is obtaining approval from the U.S. Food and Drug Administration.

“This effort truly highlights the importance of interdisciplinary collaboration,” said Angelos Barmpoutis, Ph.D., a professor at the Digital Worlds Institute at UF. “Thanks to the combined medical expertise, scientific expertise and technological expertise, we were able to accomplish a goal that will change the lives of countless individuals.”

Vaillancourt and Barmpoutis are partial owners of a company called Neuropacs whose goal is to bring this software forward, improving both patient care and clinical trials where it might be used.


Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.