Japanese researchers have conducted groundbreaking research, published in the journal “Scientific Reports,” highlighting the role of facial expressions in detecting early, subtle symptoms of depression.
The AI system used in the study demonstrated that subthreshold depression (StD), which doesn’t meet the criteria for a clinical diagnosis but is considered a risk factor for developing depression, causes subtle, unnoticeable changes in facial micro-movements. (According to DSM-5, a diagnosis of depressive disorder requires at least five out of nine symptoms, whereas in subthreshold depression, the number of symptoms is lower.)
Depression is one of the most common mental health problems, and its early symptoms often go unnoticed. This condition is frequently characterized by a reduction in facial expressiveness. However, it was previously unknown to what extent mild forms of depression, or subthreshold depression, were linked to psychomotor retardation.
Researchers from Waseda University in Japan, Professor Eriko Sugimori and doctoral student Mayu Yamaguchi, used AI to study the connection between mild forms of depression and changes in facial expressions.
The findings of this study are particularly significant because a reduction in facial expressiveness, or a “frozen face,” is also a characteristic of Parkinson’s disease. In this disease, due to a lack of dopamine, facial mimic movements are limited. Interestingly, AI has been successfully used to identify Parkinson’s disease using a similar principle.
In the study, 64 Japanese students recorded short video introductions, which were then evaluated in two ways:
- Human Evaluation: A second group of 63 students evaluated the participants’ expressiveness, friendliness, naturalness, and likability.
- AI Analysis: The same videos were analyzed using OpenFace 2.0, an AI system that examined the micro-movements of facial muscles.
The study revealed a consistent correlation: students with symptoms of subthreshold depression were rated by their peers as less friendly and expressive. However, it is noteworthy that they did not seem tense, artificial, or nervous. This suggests that subthreshold depression does not make a person openly negative but rather reduces their positive expressiveness.
The AI analysis identified specific patterns of eye and mouth movements that were more frequent in participants with subthreshold depression. These patterns included micro-movements of the eyebrows, eyelids, lips, and mouth. These subtle movements were significantly associated with the severity of depressive symptoms, even though they were unnoticeable to the human eye.
The researchers stated that their approach, based on video recordings and automated analysis of facial expressions, could be effectively used for mental health screening in schools, universities, and workplaces.
This innovative method will help in the early detection of depression (before the onset of clinical symptoms), which, in turn, will allow for timely intervention and appropriate care for individuals at risk of developing mental health problems.
