Using advanced machine learning techniques to evaluate video recordings of people engaging in basic movements, researchers have created a way to systematically measure motor functions in individuals with early-stage Parkinson’s disease and other movement disorders.
A novel approach that leverages video technology and machine learning to assess motor symptoms in the initial stages of Parkinson’s disease may facilitate the earlier identification of the condition and similar movement disorders, ultimately leading to improved treatment results.
In a recent study published in Parkinsonism and Related Disorders, researchers from the University of Florida and the Fixel Institute for Neurological Diseases demonstrate that video evaluations can assist in recognizing early signs of Parkinsonism by examining the movements of both sides of the body. This method takes advantage of the typical asymmetrical onset of Parkinson’s disease, where one side of the body is generally more affected in its early phases.
Utilizing machine learning, the researchers analyzed video recordings of participants executing simple hand and leg movements—activities frequently observed by neurologists. They focused on identifying subtle variations between healthy individuals and those with early-stage Parkinson’s disease. This technique achieved an impressive accuracy rate of 86% in differentiating between the two groups.
“Our method is non-invasive, utilizes standard video footage, and has the potential to facilitate the early detection of Parkinsonism, leading to better treatment strategies and enhanced patient care,” stated lead author Diego Guarin, an assistant professor in applied physiology and kinesiology at UF.
