An artificial intelligence (AI) tool was able to distinguish, with high accuracy, patients with Parkinson’s disease from their healthy peers by analyzing short videos of facial expressions, particularly smiles, according to a small study.
The predictive accuracy of the new tool was comparable to that of video analysis that uses motor tasks to detect Parkinson’s disease, identifying facial expressions as a potential digital biomarker for diagnosing the disease.
This type of biomarker could allow remote diagnosis without requiring personal interaction or extensive testing. This would be particularly relevant in situations such as a pandemic, with reduced mobility or in underdeveloped countries where few neurologists exist but most people have access to a phone with a camera, the researchers noted.
The study, “Facial expressions can detect Parkinson’s disease: preliminary evidence from videos collected online», was published in the form of a brief communication in the journal npj Digital Medicine.
Motor symptoms associated with Parkinson’s disease, such as tremors and muscle rigidity, affect facial muscle movements, resulting in overall reduced facial expression, also known as facial masking or hypomimia.
A growing number of studies suggest that reduced facial expressions may be an “extremely sensitive biomarker for [Parkinson’s disease]making it a promising tool for early diagnosis,” the researchers wrote.
Additionally, facial expression analysis is a non-invasive tool that only requires a webcam or phone with a camera, unlike the expensive, non-upgradable wearable sensors currently used to analyze movement during motor tasks in as digital biomarkers of Parkinson’s disease.
Now, a team of researchers from the University of Rochester, New York, has shown that analyzes of facial micro-expressions using an AI tool can accurately detect Parkinson’s disease.
The study focused on the analysis of 1,812 videos, collected online via a online tool (Park test), of 604 people (61 with Parkinson’s disease and 543 without the disease). In these videos, participants were asked to make three facial expressions – a smiling, disgusted and surprised face – each followed by a neutral face.
The average age of the participants was 63.9 years old, and most of them were Caucasian and lived in the United States. Patients with Parkinson’s disease lived with a diagnosis of the disease for an average of 8.4 years.
Changes in muscle movement for each of the three facial expressions were objectively measured and calculated in terms of nine units of action, or micro-expressions.
In line with previous research, the analysis showed that patients with Parkinson’s disease had less facial muscle movement than people without the disease. This was particularly significant for three micro-expressions: raising the cheeks and pulling the corner of the lips – usually seen when people smile – and lowering the eyebrows, usually seen when people express a disgusted face.
According to the team, “the smiling facial expression has the greatest potential to differentiate individuals with and without” Parkinson’s disease, the researchers wrote.
The team then used these differences in microexpressions to train a machine learning tool to distinguish between individuals with and without Parkinson’s disease. Machine learning is a form of AI that uses algorithms to analyze data, learn from its analyses, and then make a prediction about something.
They found that their AI tool could correctly identify patients with Parkinson’s disease based on their facial expressions with an accuracy of 95.6%, which is comparable to the prediction accuracy of 92%. reported for existing state-of-the-art video analysis that relies on limb movements.
“We show that an algorithm’s ability to analyze subtle features of facial expressions, often invisible to the naked eye, adds important new insights to a neurologist,” the team wrote.
Thus, facial expressions, in particular the smile, “may become a reliable biomarker for [Parkinson’s disease] detection,” they added.
This type of diagnostic biomarker does not require in-person examination and diagnosis, which can be “potentially transformative for patients who require remote diagnostics due to physical separation (e.g., due to COVID) or stillness,” the team wrote.
Additionally, “many underdeveloped regions and underrepresented populations can also benefit from such a method that uses facial expressions without worrying about direct access to a neurologist,” the researchers added.
They noted that in 23 African countries, the average population per neurologist exceeds half a million, but that about 75% of Africans have access to a mobile phone, and that similar situations exist in Asia and South America.
Additionally, the authors hypothesize the potential value of an app that, with the user’s permission, could automatically analyze the user’s smile micro-expressions (from a short video) and provide a referral in case of signs of Parkinson’s risk.