Artificial intelligence-powered video processing technology developed at the University of Florida could help neurologists more accurately track the progression of Parkinson's disease in patients, ultimately improving patient care and quality of life.
The system, developed by Dr. Diego Guarin, assistant professor of applied physiology and kinesiology in the College of Health and Human Performance at the University of Florida, applies machine learning to analyze video recordings of patients performing the finger tapping test (rapidly tapping the thumb and index finger 10 times), a standard test for Parkinson's disease.
“By studying these videos, we can detect even the most subtle changes in hand movements that are characteristic of Parkinson's disease but that may be difficult for clinicians to visually identify,” said Guarin, who is in the Norman Fixel Institute for Neurological Disorders at the University of Florida Health System. “The beauty of this technology is that patients can record themselves performing the test, and the software analyzes it and reports the patient's movements to the clinician so they can make a decision.”
Parkinson's disease is a brain disorder that affects movement, causing slowness, tremors, stiffness, and difficulties with balance and coordination. Symptoms usually appear gradually and worsen over time. There is no specific test or imaging that can diagnose Parkinson's disease, but a range of movements and tasks performed by patients can help clinicians identify and assess the severity of the disorder.
The rating scale most commonly used to track the progress of Parkinson's disease is the Movement Disorder Society's Unified Parkinson's Disease Rating Scale. Guarin explained that although it is reliable, the rating is limited to a five-point scale, limiting the ability to track subtle changes in progression and making it prone to subjective interpretation.
A research team including University of Florida neurologists Joshua Wong, MD, Nicholas McFarland, MD, and Adolfo Ramirez Zamora, MD, has devised a more objective way to quantify motor symptoms in people with Parkinson's disease by using machine learning algorithms to analyze videos and capture subtle changes in the disease over time.
“We found that with a camera and a computer, we can observe the same characteristics that a clinician would look at,” Guarin says. “With the help of AI, the same test can be performed easier and faster for all involved.” Guarin said the automated system also used the precise data collected by the camera to reveal previously unnoticed details about movements, such as how quickly a patient opens and closes their fingers during a movement and how the movement characteristics change with each tap.
“We found that people with Parkinson's disease have a delayed opening movement compared to the same movement in healthy people,” Guarin said. “This is new information that would be nearly impossible to measure without video and computers, and it shows that this technology can provide a more accurate picture of how Parkinson's affects movement and provide new markers to help evaluate the effectiveness of treatment.”
To perfect the system that Guarin originally designed to analyze facial features for conditions other than Parkinson's, the team used the University of Florida's HiPerGator, one of the world's largest AI supercomputers, to train several models.
“HiPerGator helped us develop machine learning models that simplify video data into behavior scores,” Guarin explains. “We used HiPerGator to train, test, and refine different models using large amounts of video data, and now we can run those models on smartphones.”
Michael S. Okun, MD, director of the Norman Fixel Research Institute and medical advisor to the Parkinson's Disease Foundation, said automated, video-based assessments could be a “game-changer” in clinical trials and treatment.
“The finger tapping test is one of the most important components used to diagnose Parkinson's disease and measure disease progression,” Okun said. “Currently, experts are needed to interpret the results, but it is a breakthrough that Diego and three other Parkinson's neurologists from Fixel Institute have been able to use AI to objectify disease progression.”
In addition to providing the technology to neurologists and other healthcare providers, Guarin is also working with UFIT to develop an app for mobile devices so individuals can assess their disease over time at home.
July 23, 2024
