It can be challenging to accurately evaluate the seriousness of an individual’s condition when they are grappling with significant depression. In the near future, an AI-driven facial analysis mobile application might offer insights into a person’s well-being.
Known as MoodCapture, this innovative program is presently in development by a team of researchers at Dartmouth College in New Hampshire.
The application operates by capturing multiple photos of the user’s face and surroundings through the back camera each time the phone is unlocked using facial recognition technology. These images are then scrutinized by an AI algorithm that assesses the individual’s facial expressions and the background details.
If the algorithm detects a deterioration in a patient’s depression levels, it will recommend activities like socializing with loved ones or engaging in outdoor exercises. Importantly, the app is designed not to immediately suggest seeking psychiatric help, as this could potentially exacerbate the individual’s feelings of despair.
A group of 177 participants, segmented into five subsets, all diagnosed with major depressive disorder, underwent training by the AI system.
Over a span of 90 days, each individual’s mobile device captured photos while they responded to the statement, “I have felt down, sad, or hopeless,” as part of the eight-point Patient Health Questionnaire commonly used to gauge despair.
Despite consenting to the photo-taking process, the participants were unaware of when the images were being captured until they acknowledged the prompt. This is a critical aspect as it ensures that their emotions were not being consciously masked during the photo-taking moments.
Upon analyzing a total of 125,000 photos, the AI identified specific facial expressions (in certain subsets) that correlated most strongly with the aforementioned statement. These expressions included variations in eyebrow positioning, eye movements, head orientation, and muscle tension. Furthermore, recurring environmental factors such as lighting conditions and the presence of other individuals were also noted.
Subsequently, the developed AI model was utilized to evaluate images of the remaining subsets. The software demonstrated a 75% accuracy rate in identifying individuals experiencing worsening despair. With further refinement, it is anticipated that the accuracy will reach at least 90% within the next five years.
MoodCapture’s primary advantage lies in enabling patients to monitor their condition more frequently, allowing for prompt intervention during downturns, in contrast to traditional psychiatric assessments which may offer similar insights but at a slower pace.
Professor Nicholas Jacobson, a co-author of the study, highlighted, “This approach acknowledges the episodic and fluctuating nature of major depressive disorder, where symptoms can vary significantly on a daily basis.” The research aimed to identify specific patterns and characteristics indicative of changes in despair over time among individuals with MDD.
The research originates from Dartmouth College.
