AI-powered MRI accurately distinguishes Parkinson’s from similar disorders

Credit: Unsplash+

A team of researchers from the University of Florida has developed a powerful new tool that uses artificial intelligence (AI) and MRI scans to distinguish Parkinson’s disease (PD) from other similar neurological disorders with remarkable accuracy.

The method, known as Automated Imaging Differentiation for Parkinsonism (AIDP), could significantly improve how doctors diagnose and treat people with movement disorders.

Parkinson’s disease shares many symptoms with other conditions, such as multiple system atrophy (MSA) and progressive supranuclear palsy (PSP).

These disorders are known as atypical parkinsonian syndromes and often present with overlapping symptoms like stiffness, tremors, and difficulty with movement.

This similarity makes it difficult to tell them apart, especially in the early stages. Even experienced doctors using traditional methods, such as physical exams and standard imaging, often misdiagnose these conditions.

While certain tests like dopamine transporter imaging have been approved since 2011 to help distinguish Parkinson’s from conditions like essential tremor, they still struggle to tell the difference between PD, MSA, and PSP. Other approaches, like skin biopsies and tests for protein build-up in the brain, have also been limited—especially in diagnosing PSP accurately.

In this new study, published in JAMA Neurology, researchers tested AIDP, a machine-learning model that examines MRI scans to detect tiny changes in brain structure that differ between these conditions.

The study included data from 249 patients enrolled across 21 clinical sites in the U.S. and Canada, along with an additional 396 cases used to train the model. All patients had confirmed diagnoses of PD, MSA, or PSP.

AIDP works by analyzing advanced diffusion MRI scans, which track how water moves through the brain. This method reveals structural changes and patterns of brain degeneration specific to each disease. Researchers looked at 132 different areas of the brain—including the cortex, brainstem, and cerebellum—to teach the AI model how to recognize disease-specific features.

Key Results:

  • AIDP correctly identified Parkinson’s disease over atypical cases 96% of the time.
  • It distinguished MSA from PSP with 98% accuracy.
  • It achieved 98% accuracy when separating Parkinson’s from either MSA or PSP.
  • When compared to diagnoses confirmed by autopsy (the gold standard), AIDP achieved an overall accuracy of 93.9%, which is 12.3% better than traditional clinical diagnoses.

These results suggest AIDP is not only more accurate than current methods, but also more consistent, even in tough-to-diagnose cases. Its use of MRI makes it non-invasive, and it works across different imaging centers and machines—making it easier to adopt in various hospitals and clinics.

One of the most promising aspects of AIDP is that it performs better than the commonly used dopamine transporter imaging, which cannot clearly distinguish between PD, MSA, and PSP. By contrast, AIDP pinpoints clear structural differences between these disorders.

AIDP could soon become a valuable addition to routine medical practice for patients with suspected Parkinsonian disorders. It could help patients get the right diagnosis earlier, leading to more targeted treatments and better outcomes.

Researchers also suggest combining AIDP with other diagnostic tools—such as tests for protein markers in the brain or skin biopsies—to increase diagnostic confidence even further. With such integration, doctors could offer faster and more precise answers to patients and their families.

This study marks a major step forward in using AI for neurological diagnoses. Many patients with Parkinson’s-like symptoms spend years without a clear answer, which delays proper treatment. AIDP offers a practical, scalable solution that could cut through the uncertainty.

Its strength lies in using real patient data from multiple clinical centers and showing improved accuracy even against autopsy results. The study also highlights how AI can complement—not replace—clinicians by providing more accurate tools for decision-making.

Overall, AIDP represents a significant leap toward precision medicine in neurology. If adopted widely, it could reduce misdiagnosis rates, improve patient care, and serve as a model for using machine learning to solve other complex medical problems.

If you care about Parkinson’s disease, please read studies that Vitamin B may slow down cognitive decline, and Mediterranean diet could help lower risk of Parkinson’s.

For more health information, please see recent studies about how wheat gluten might be influencing our brain health, and Olive oil: a daily dose for better brain health..

The research findings can be found in JAMA Neurology.

Copyright © 2025 Knowridge Science Report. All rights reserved.