Just as our emotions control our the body, determining our posture and expression and tensions; our bodies are our feeling devices, with the sensory input and muscle tension in turn causing emotions. The two work in tandem to create our experience and expression, meaning maintaining muscle control is vital for connection with the world and those around us.
In Japan, researchers wanted to see how Parkinson’s patients had their emotions and perceived mood effected as they progressively lose control of their facial muscles.
Using facial recognition software, researchers found the emotional expression was effected in patients with Parkinson’s. In the study “Detecting facial characteristics of Parkinson’s disease by novel artificial intelligence (AI) softwares,” researchers used facial recognition software to predict a person’s age and mood, in order to determine how impaired motor skills impacts emotions. They contend the person’s inability to use their facial muscles to express their full range of emotions has implications on their ability emotional health, self esteem, and relationships.
Parkinson’s disease is a rare neurological disorder that progressively erodes a person’s control of their movement. Usually emerging later in life, patients begin to notice a slight tremble or slowness, over time progressing to muscle rigidity, loss of balance, and slowed movement. Eventually they lose so much muscle control they cannot continue to live independently. The disease typically begins later in life, with most diagnosis occurring after age 50. While Parkinson’s is a relatively common and well-documented disease, the true underlying cause and nature of the disease is still not understood, and treatment options are still symptomatic.
The New Tech
The team employed artificial intelligence that can pick up on changes to facial characteristics at greater depth and precision. The new AI technology can pick up on more details such as changes in the texture of the skin, emotions, age, and tension.
They evaluated their AI’s findings against tradition methods of facial detection and the algorithmic analysis. They wanted to see if their ever smarter technology might find details earlier methods have missed.
Lead author of the study, Koh Tadokoro, MD said in a press release,
“Whereas most previous studies evaluated the facial changes caused by PD during movement or tasks, we found evidence of specific facial changes based solely on a single photograph using modern AI.”
They expect this to be a step up from medley of earlier methods employed to gather data— which gathered data from a patient’s webcam. Their application requires patients to use apps including the “Face Log” app and Microsoft Azure Face.
The team at Okayama University studied the facial expressions of 193 participants, half with Parkinson’s and half as a control group. They first took just a single picture of each subject, and the AI and the Azure Face software each scanned the faces using use algorithms that read muscle tension, comparing the control and patient group for predicted age and emotion. The participants each had a single photo taken for analysis, without any instructions about expression.
The machines indeed picked up on the physical signals of wear and tear, with Parkinson’s patients appearing older and both having less emotional expression and appearing overall less happy.
While the algorithm has a bias in underestimating age for Asian populations, once adjusted for the Japanese population, they found that while the AI generally corrected predicted the ages of those in the control group, it tended to overestimate the age of those with Parkinson’s by an average of 2.4 years. The difference was even greater for patients who were younger or male.
The AI also observed that the Parkinson’s patients had their expressions muted, with a higher score (88.9% vs a control group of 76.6%) on the quality of appearing “mask like”, a quality associated with Parkinson’s due to their symptomatic slowness of movement, bradykinesia. Since our muscles hold so much of our emotions and the places we hold tension define so much of our expression, patients struggle to feel they can express emotions at the same depth, and similarly, without the physical manifestation of the emotion, cannot feel it as vividly either.
The participants were also analyzed for happiness, and similarly, only a fraction of Parkinson’s patients were observed as happy (4.7%) compared to the control (18.5%). The tech didn’t find any differences in their analysis of the skin, which searched for variations in wrinkles, pores, skin tone, and dark rings under the eyes.
The doctors emphasized the importance of facial expression for social outcomes, first impressions, and one’s own mental health. Without control of ones face, patient’s often feel out of control in their communication too.
Accurate for Who?
Still, the doctor’s also wanted to call attention to the downsides of the novelty of AI’s facial recognition, noting that there are ethical concerns regarding it’s people of color and women. In the study, they wrote,
“One is that classification accuracy of commercial facial recognition software differs depending on gender and skin color, which was seriously lower for darker female people compared with lighter male.”
They note that they even ran up against their barrier in their own study, requiring them to adjust the age predictions for the Asian population.
With adjustments to the software, the doctors expect the AI become a speedy friend of the Parkinson’s community, serving both as a diagnostic tool and as part of treatment.