Depression remains a global mental health concern, and early detection is essential.
A new study published in Scientific Reports examined how individuals with mild depressive symptoms differ from their peers in facial expression and how they are perceived by others.
“We aimed to determine whether subtle facial cues associated with early depressive tendencies could be detected through both subjective impression ratings and automated facial analysis,” study author Eriko Sugimori told us. “We developed three main hypotheses.”
The research team examined how individuals with subthreshold depression would show reduced positive facial expressions and distinct patterns in eye and mouth movement. They also looked at how these individuals would be rated as less expressive, natural, friendly, and likable by others, and how the raters’ own depressive tendencies would not significantly bias their evaluations.
“These hypotheses informed our design and analysis approach,” Sugimori told us. “I have long been interested in non-verbal indicators such as facial expressions that may reveal psychological states before clinical symptoms fully emerge. Additionally, I was curious about how cultural norms in East Asia, where emotional expression is often more restrained, might influence these early signs.”
The research team recorded ten-second self-introduction videos of Japanese university students. These were then rated by a separate group of peer evaluators on various impression traits. At the same time, the researchers used OpenFace 2.0, an AI-based facial behavior analysis tool to extract specific muscle movements. They then statistically analyzed how these subjective and objective data correlated with the participants’ depression scores.
“Individuals with subthreshold depression received significantly lower scores on positive impression traits such as expressiveness, naturalness, friendliness, and likability,” Sugimori told us. “Notably, there were no significant differences in negative traits like stiffness or nervousness. »
Objectively, facial muscle movements around the eyes and mouth such as AU01 (inner brow raiser) and AU26 (jaw drop) were more pronounced and correlated with depressive symptom severity.
“One surprising and reassuring finding was that raters’ own mood states did not affect their evaluations, suggesting the impressions were shaped by the expressions of the subjects themselves,” Sugimori told us. “Also, the fact that only positive traits declined, while negative ones remained stable, provided nuanced insight into how subthreshold depression may manifest socially.”
The researchers believe their findings support the idea that subtle facial dynamics may serve as early, non-invasive markers of depression risk, possibly even before clinical diagnosis.
“In the future, this could inform low-cost screening tools in educational or occupational settings,” Sugimori told us. “However, it’s essential to incorporate cultural sensitivity, as baseline expressivity norms differ globally. I would be very interested to see this line of research expand through deep learning approaches and with more diverse cultural samples. Such efforts could help build AI tools that not only detect early signs of mental health issues, but also foster better understanding and destigmatization of emotional expression across cultures.”