When someone is suffering from major depression, objectively self-gauging the severity of their condition can be challenging. Well, within just a few years, an artificially intelligent face-analyzing smartphone app may tell such folks how they're holding up.
Currently being developed by a team of scientists at New Hampshire's Dartmouth College, the prototype application is known as MoodCapture.
The idea behind the app is that whenever the user unlocks their phone via its facial recognition system, the device's front camera will capture multiple photos of their face and surroundings. An AI-based algorithm will then assess those images, scrutinizing the user's facial expression along with background images.
If the app determines that the user's depression is worsening, it will suggest measures such as getting outdoor exercise or socializing with family and friends. Ideally it won't sound a stern warning that the user seek psychiatric attention – at least, not at first – as doing so might just make the person feel worse about their situation, thus strengthening their depression.
The AI was trained on a group of 177 test subjects divided into five subsets, all of whom had previously been diagnosed with major depressive disorder.
Over a period of 90 days, each person's phone shot photos of them as they rated how much they agreed with the statement, "I have felt down, depressed, or hopeless." That prompt is part of the eight-point Patient Health Questionnaire, which is widely used to assess depression.
Although the participants had agreed to having their picture taken by their phone, they weren't aware that it was doing so as they responded to the prompt. This is an important consideration, as it means they weren't subconsciously masking their emotions when the photos were taken.
When the total of 125,000 photos were subsequently analyzed, the AI identified the facial expressions (in some subsets) which coincided with the most emphatic agreements to the prompt. Such expressions included variations in gaze direction, eye movement, positioning of the head, and muscle rigidity. The AI also identified recurring environmental factors, such as bright or dim lighting and the presence or absence of other people.
Utilizing the resulting AI model, the app was next used to analyze smartphone images of the other subsets. The app proved to be 75% accurate at identifying which people were experiencing worsening depression. It is believed that once the technology is developed further – within about five years – the accuracy rate should climb to at least 90%.
And while periodic clinical psychiatric assessments might provide the same basic information, the big advantage of MoodCapture is that it should allow patients to assess their illness much more frequently, quickly responding to downswings before they progress too far.
"This method acknowledges the dynamic and highly individualized nature of MDD [major depressive disorder], where symptoms can change significantly from day to day," study co-author Prof. Nicholas Jacobson told us. "By tracking these changes closely in a group of individuals diagnosed with MDD, we aimed to uncover patterns and characteristics specific to changes in depression over time."
Source: Dartmouth College