Facial Recognition Program IDs Parkinson’s Patients as Older, Expressionless

Facial Recognition Program IDs Parkinson’s Patients as Older, Expressionless
3.3
(7)

Differences in the facial appearance and emotional expressions of people with Parkinson’s disease and  healthy adults were evident to a facial recognition software program, researchers in Japan reported.

The program’s algorithm consistently overestimated the age of the nearly 100 patients in this study by a few years, particularly for males, and perceived their expressions as “emotionless” significantly more than those of the age-matched adults serving as a control group.

With improvements, especially those that raise ethical concerns like a greater inaccuracy with darker skin tones, such programs may help in diagnosing and managing Parkinson’s, the researchers wrote.

Their study, “Detecting facial characteristics of Parkinson’s disease by novel artificial intelligence (AI) softwares,” was published in the journal Brain Supplement.

Parkinson’s-associated motor symptoms, such as tremors and muscular rigidity, affect a person’s ability to show emotions via facial expressions, with implications for a patient’s sense of self-esteem and engagement in social life.

Artificial intelligence (AI) may be more able than other methods to track and quantify these changes with disease progression.

AI-based facial recognition technology analyzes facial characteristics, such as age, emotions, and skin texture, and could prove useful in assessing Parkinson’s-related changes.

But the “classification accuracy of commercial facial recognition software differs depending on gender and skin color,” the scientists wrote, and this “potential ethical issue should be carefully discussed and resolved to apply the facial recognition technology to medical situations.”

Researchers at Okayama University recruited 97 adults with Parkinson’s — at different disease stages — and 96 healthy adults (mean age of 69.5 for both groups) to their study.

“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,” Koh Tadokoro, MD, the study’s lead author, said in a university press release.

Tadokoro and his colleagues took a snapshot of each participant’s face using a webcam, which they then analyzed for age, gender, and emotion, using the commercially available program Microsoft Azure Face. Skin texture was analyzed using the “Face Log” smartphone app (v1.08), adjusted for a Japanese population.

Participants received no instructions regarding facial expression while being photographed.

Azure Face identified differences in people’s apparent age — the difference between AI-calculated age and actual age — and emotion. Age had to be adjusted for an Asian population, which were initially underestimated.

The algorithm then tended to score people with Parkinson’s as older than they really were, while ages for those in the control groups were generally accurate. The age gap for Parkinson’s patients was about 2.4 years and zero for controls. This effect appeared larger for men and younger patients, than for women and older patients.

Patients’ faces were also scored as expressionless or “mask like” more frequently than those of controls (88.9% vs .76.6%, respectively, on average) and less frequently as happy (4.7% vs 18.5%). Mask-like expressions are known in Parkinson’s patients and attributed to bradykinesia, or slow voluntary movement.

No differences between the two groups were found in the skin analysis, which looked at things like wrinkles, shadows under the eyes, and pore texture.

Biases in skin tone and gender-related issues, as well as the tendency for commercial software to underestimate ages of Asians, must be resolved before it can be used in clinical settings. Nonetheless, the researchers see potential in its future applications.

“We detected that [Parkinson’s] patients looked older and expressionless using [publicly] available AI face recognition software,” they concluded.

“AI facial recognition is an innovative and powerful technology,” they added, with potential for clinical use if ethical issues are adequately addressed.

“We hope our study accelerates the use of AI technology for the diagnosis and treatment of patients with Parkinson’s and other neurodegenerative diseases,” Tadokoro said.

Forest Ray received his PhD in systems biology from Columbia University, where he developed tools to match drug side effects to other diseases. He has since worked as a journalist and science writer, covering topics from rare diseases to the intersection between environmental science and social justice. He currently lives in Long Beach, California.
Total Posts: 208
Ana holds a PhD in Immunology from the University of Lisbon and worked as a postdoctoral researcher at Instituto de Medicina Molecular (iMM) in Lisbon, Portugal. She graduated with a BSc in Genetics from the University of Newcastle and received a Masters in Biomolecular Archaeology from the University of Manchester, England. After leaving the lab to pursue a career in Science Communication, she served as the Director of Science Communication at iMM.
×
Forest Ray received his PhD in systems biology from Columbia University, where he developed tools to match drug side effects to other diseases. He has since worked as a journalist and science writer, covering topics from rare diseases to the intersection between environmental science and social justice. He currently lives in Long Beach, California.
Latest Posts
  • walking ability, music-based digital therapy
  • rehabilitation study
  • parkin protein and the brain
  • movements

How useful was this post?

Click on a star to rate it!

Average rating 3.3 / 5. Vote count: 7

No votes so far! Be the first to rate this post.

As you found this post useful...

Follow us on social media!

We are sorry that this post was not useful for you!

Let us improve this post!

Tell us how we can improve this post?