Video Selfies May One Day Help in Diagnosing Parkinson’s
A computer software that tracks how facial muscles move during a video selfie may be able to help predict whether a person is likely to develop Parkinson’s disease, a study reported.
The study involved the use of the PARK test, a web-based application designed to detect and objectively measure symptoms of Parkinson’s through video analytics. The application could one day assist people anywhere in getting a diagnosis — as long as they have a webcam, a computer, and an internet connection.
“Objective, digital assessments of Parkinson’s disease can help us diagnose people with the condition and evaluate new therapies for the condition faster,” Earl Ray Dorsey, MD, one of the study’s authors, said in a press release.
The study, “Facial expressions can detect Parkinson’s disease: preliminary evidence from videos collected online,” was published in npj Digital Medicine by a team of researchers at the University of Rochester, which developed the test.
One of the motor symptoms of Parkinson’s is increased stiffness of facial muscles. Because the facial muscles that control expression do not move as much as they normally would, the face becomes less expressive. This is called hypomimia or facial masking.
Many people these days take pictures or record videos of themselves with a smartphone or webcam, and make facial expressions when doing so.
“What if, with people’s permission, we could analyze those selfies and give them a referral in case they are showing early signs [of Parkinson’s]?,” Ehsan Hoque, PhD, another study’s author, asked.
To answer this question, the researchers combined computer vision software with machine learning to detect from digital pictures or videos subtle facial movements (micro-expressions) that are otherwise invisible to the naked eye.
The team watched 1,812 videos featuring 604 people — 543 without Parkinson’s and 61 with Parkinson’s — with a mean age of 63.9 years who were asked to complete a series of voice and motor tasks online while being recorded on a webcam.
The tasks, which generally take 20 minutes to complete, included reading a complex sentence aloud, tapping the index finger to the thumb 10 times as quickly as possible, and making three different facial expressions (smiling, and a disgusted and surprised face) three times, alternating each expression with a neutral face.
Compared with people without the disease, those with Parkinson’s made fewer movements with facial muscles, particularly those involved in raising the cheeks, pulling the lip corners, and lowering the eyebrows.
“The smiling facial expression has the greatest potential in differentiating individuals with and without [Parkinson’s],” the researchers wrote.
Detection of facial muscle movement was about as accurate as other methods of video analysis focusing on limb movements (95% vs. 93%). The output is a percentage likelihood of developing Parkinson’s.
Not all people with Parkinson’s show each and every symptom of the disease, the researchers noted, which means that its identification should not rely on a single diagnostic modality.
“Our findings only show that facial expressions, especially smiling, can be used as one of the reliable modalities,” they wrote.
One advantage of using a web-based application is that “patients do not need to be near a neurologist for an in-person diagnosis,” the researchers wrote. “This is potentially transformative for patients in need of physical separation from others (e.g., due to COVID-19) or are immobile.”
“Additionally, this can enable low-cost screening where access to a neurologist is limited,” they added.
While the researchers have made progress in detecting Parkinson’s by automatically tracking how facial muscles move, further work is needed before requesting U.S. Food and Drug Administration permission to analyze people’s video selfies.
For now, the Rochester team is advancing its research with a $500,000 grant from the Gordon and Betty Moore Foundation.
“The foundation wants us to validate the feedback that we would give people if they did, indeed, show early signs of Parkinson’s — especially if they are performing the test at home,” Hoque said. “The challenge is not only validating the accuracy of our algorithms but also translating the raw machine-generated output in[to] a language that is humane, assuring, understandable, and empowering to the patients.”
The team noted that several other movement disorders — such as ataxia, Huntington’s disease, and multiple dystrophy — can share some features with Parkinson’s, including involuntary tremors, and is trying to address that.
“We are in a pursuit of differentiating those tremors using AI [artificial intelligence] to prevent the potential harm of misdiagnosis while maximizing benefit,” Hoque added.