AI App Assesses Depression Through Facial Analysis

Measuring the severity of major depression can be difficult for individuals experiencing it. However, in the near future, a smartphone app utilizing artificial intelligence to analyze facial expressions may provide insight into one’s condition.

Developed by scientists at Dartmouth College in New Hampshire, the prototype app, called MoodCapture, aims to fulfill this purpose.

The concept behind the app involves utilizing the facial recognition system each time the user unlocks their phone. Upon activation, the front camera captures multiple images of the user’s face and surroundings. An AI algorithm then analyzes these images, focusing on the user’s facial expression and background details.

App Recommendations and Approach

Should the app detect a deterioration in the user’s depression, it will recommend actions like outdoor exercise or socializing with loved ones. Initially, it aims not to issue a stern warning for psychiatric attention, as this could potentially exacerbate the individual’s feelings and reinforce their depression.

The AI underwent training using a cohort of 177 individuals, all diagnosed with major depressive disorder, split into five subsets.

Over a 90-day period, participants rated their agreement with the statement “I have felt down, depressed, or hopeless” while their phones captured photos of them. This statement is part of the widely used eight-point Patient Health Questionnaire for assessing depression.

Importantly, participants were unaware that their phones were taking their pictures while they responded to the prompt, ensuring their emotions were not subconsciously concealed during the photo capture process.

Facial Expression Analysis

After analyzing a total of 125,000 photos, the AI detected facial expressions (in certain subsets) that correlated with strong agreement to the prompt. These expressions included variations in gaze direction, eye movement, head positioning, and muscle tension. Additionally, recurring environmental factors such as lighting conditions and the presence of others were identified.

Using the AI model developed from this analysis, the app then evaluated smartphone images of the remaining subsets. It achieved a 75% accuracy rate in identifying individuals experiencing worsening depression. Researchers anticipate that with further technological advancements in approximately five years, this accuracy rate will increase to at least 90%.

Unlike periodic clinical psychiatric assessments, MoodCapture offers the advantage of allowing patients to assess their condition more frequently, enabling prompt intervention during downturns before they escalate.

Co-author of the study, Prof. Nicholas Jacobson, highlighted the dynamic and individualized nature of major depressive disorder (MDD). By closely monitoring changes in symptoms among individuals diagnosed with MDD, the study aimed to identify specific patterns and characteristics associated with fluctuations in depression over time.

Read the article on: New Atlas

Read more: Pair Mindfulness with Exercise for a Mental Health Boost in 2024