Researchers from Shenzhen University and Fudan University (China) have found that excessive influence of past experience on face perception may be an early sign of a predisposition to depression. The work was published in the journal Science Advances.
The study involved 1,332 adolescents from the European population cohort IMAGEN. The scientists analyzed functional magnetic resonance imaging (fMRI - a method that measures blood flow in the brain to identify active areas) data while participants watched short videos in which a neutral facial expression turned angry. Using a machine learning algorithm, they extracted a mathematical representation of how the visual cortex encodes angry faces and called this pattern a "visual code."
Participants were divided into two groups: those with typical efficient encoding and those with inefficient, "noisy" processing. Adolescents in the second group reported higher levels of emotional symptoms - anxiety and depressed mood - on clinical questionnaires. This association indicated that distorted perception of angry faces is linked to emotional distress. However, correlation alone does not prove causation.
To understand the mechanism, the researchers built a deep neural network - an artificial intelligence that mimics the workings of the brain. The model had two pathways: one processed visual information, the other stored abstract knowledge about emotions. Normally, abstract knowledge helps recognize facial expressions faster. When the scientists strengthened the influence of the conceptual pathway on the visual one, the model began to misclassify neutral faces as angry or sad. This effect of excessive reliance on expectations was called "visual over-regulation." The internal code of such a model matched the "noisy" brain patterns of adolescents with high levels of emotional symptoms.
The researchers then created a quantitative marker of the degree of visual over-regulation based on brain scans. Four years later, they examined 725 of these participants. Among those who initially had no clinical symptoms, the marker successfully predicted who would develop pronounced emotional problems. The prediction proved more accurate than traditional measures based only on demographics and baseline brain activity.
A specificity check showed that the marker is linked to a genetic variant associated with major depressive disorder and to polygenic risk for depression (a summary estimate of the influence of many genes on the likelihood of illness). No links were found with genetic risk for alcohol dependence or eating disorders.
In a separate clinical sample of 134 patients diagnosed with depression and 62 healthy people, the marker was higher in patients and accurately distinguished the groups. In alcohol dependence, anorexia and bulimia, no such differences were found.
The authors note limitations: participants were predominantly white people of middle and high income, and differences in symptoms between groups with different processing efficiency turned out to be small. This means that at this stage the marker alone has limited diagnostic value for an individual patient. Further validation in independent groups is needed. It also remains to be determined whether the perceptual distortion is a stable trait or a temporary state, and whether methods that "retrain" visual perception can reduce the risk of depression.