A multicenter research project led by Deepak Gupta, M.D., M.S., associate professor of neurological sciences at the University of Vermont (UVM) Larner College of Medicine, has led to this novel tool, which was announced in January in the peer-reviewed journal Scientific Reports. The tool uses an artificial intelligence (AI) machine learning informatics system to support a physician’s clinical decision‑making in real time during a patient’s visit.
“Parkinson’s disease affects movement and can cause tremor, slowness, stiffness, and changes in walking speed and balance. Because symptoms and signs can overlap with other conditions early in the disease course, an early and accurate diagnosis can be challenging,” says Dr. Gupta, who is also the director of clinical informatics for neurology at UVM Health and a movement disorders neurologist at the UVM Medical Center.
Parkinson’s is the second most common neurodegenerative disorder, after Alzheimer’s disease, and is the fastest-growing neurological disorder globally. There is an increasing concern in the health care community that this growing worldwide prevalence points to a potential Parkinson’s epidemic. Therefore, an early and accurate diagnosis of Parkinson’s is crucial for both clinical care and research.
The new tool, called Clinical Decision Support platform for Parkinson’s Disease (CDS-PD), integrates information collected during a patient’s visit with a Parkinson’s specialist with standard clinical diagnostic criteria for Parkinson’s disease to classify the patient’s clinical diagnosis in real time at point of care. Additionally, CDS-PD use machine learning techniques to perform disease subtyping and computes prognosis measures based on commonly collected clinical scales.
“Artificial intelligence–driven clinical informatics tools like CDS-PD have the potential to bridge the gap between availability of diagnostic criteria and clinical practice guidelines to improve the accuracy of diagnoses,” Gupta says.
Multiple Causes, Increasing Risk
Parkinson’s disease occurs when the brain cells that make dopamine, a chemical that coordinates movement, gradually fail and die, causing people to experience tremor, slowness, walking difficulty, and imbalance. Parkinson’s disease is progressive, worsening slowly over time, and there is no known cure. Most cases are idiopathic, meaning that the cause is unknown. Specific gene mutations are linked to hereditary cases. Long-term exposure to neurotoxins in solvents, pesticides, and herbicides including Agent Orange is associated with increased risk of developing Parkinson’s disease. Traumatic brain injury has also been linked to an elevated risk.
Such complexities and growing worldwide prevalence highlight the critical need to develop innovative approaches for improving diagnostic and prognostic assessment of Parkinson’s, including comparison between neurotoxin-associated and idiopathic Parkinson’s disease. For example, Gupta and his colleagues have used the CDS-PD to show for the first time in a prospective study of U.S. Veterans that Parkinson’s patients with exposure to Agent Orange might have lower cognitive performance and greater patient-reported motor disability compared to patients without such exposure.
“Because symptoms and signs can overlap with other conditions early in the disease course, an early and accurate diagnosis can be challenging … Artificial intelligence–driven clinical informatics tools like CDS-PD have the potential to improve the accuracy of diagnoses.” — Deepak Gupta, M.D., M.S.
Gupta and his collaborators are currently expanding the CDS-PD to investigate AI-based methods of studying cognitive decline in Parkinson’s disease and to implement a practical diagnostic algorithm to help general neurologists in differentiating between Parkinson’s disease and related atypical parkinsonian disorders. In future, they plan to deploy the latest technology frameworks to make CDS-PD interoperable with electronic health records systems and integrate with Ambient AI in clinical practice.
Development of CDS‑PD has been supported by the U.S. Department of Defense through an Early Investigator Research Award in 2021 and Investigator Initiated Research Award in 2023. The co-authors from UVM on this scholarly publication include Larner Class of 2026 medical student Cole Zweber, who recently matched for neurology residency at Stanford University; UVM Visiting Scholar Vivikta Iyer, M.D., who recently matched for neurology residency at the University of Arkansas; Ian Zurlo, who worked on this project as an undergraduate student under a Laud Student Fellowship provided by the Frederick C. Binter Center for Parkinson’s Disease & Movement Disorders at UVM Medical Center; and James T. Boyd, M.D., the Robert W. Hamill, M.D., Green and Gold Professor of Neurological Sciences and director of the Binter Center for Parkinson’s Disease & Movement Disorders. Collaborating institutions included Case Western Reserve University in Cleveland, Oregon Health Sciences University, and the VA Health Care System in Portland, Oregon.
Read the full publication in Scientific Reports
Research like this has contributed to the University of Vermont’s designation by the Carnegie Classification of Institutions of Higher Education as an R1 institution, placing it in the top tier of research universities in the U.S.
