Researchers at Dartmouth University have trained the AI application MoodCapture to diagnose depression before symptoms appear.
The app uses the phone’s front camera to scan emotions.
“The app conducts an instant analysis as you use the facial unlock feature throughout the day. If we address the complex issues of privacy and ethics, MoodCapture could significantly aid people with depression in the future,” said one of the developers, Andrew Campbell.
The team has published a preliminary version of their work and plans to present full results at the Association for Computing Machinery conference in May 2024.
The study involved 177 individuals diagnosed with major depressive disorder, who used the app for three months. Over 125,000 images were collected.
MoodCapture successfully identified early symptoms of depression in users with 75% accuracy. The team also developed a more personalized model, achieving an 80% success rate.
“The neural network focuses on facial features, expressions, and gaze. It extracts meaningful information from each image, including lighting conditions, the number of people in the shot, dominant colors, location, and background objects,” Campbell added.
The authors noted that the app can detect signs of depression before the user realizes there is a problem, potentially allowing for treatment before the condition worsens.
Earlier in February, scientists at the University of California discovered a method using AI to predict Alzheimer’s disease seven years before symptoms appear.