Medical News Observer

Key Points:

• MoodCapture AI app uses the phone’s front camera to capture images for depression monitoring. 
• AI model analyzed over 125,000 images for depression indicators, achieving a 75% accuracy rate in early symptom detection.
• The app’s AI model correlates specific facial expressions and environmental surroundings with depression, aiming to refine accuracy to 90% through further model tuning with patient input.

AI-Powered MoodCapture App

Researchers at Dartmouth have developed an innovative smartphone application named MoodCapture that employs artificial intelligence alongside facial-image processing to detect signs of depression proactively. Utilizing the phone’s front camera, MoodCapture captures and analyzes a user’s facial expressions and environmental surroundings during regular phone use, seeking out clinical indicators linked with depression. 

Clinical Study

During a significant research study spanning 90 days, 177 participants diagnosed with major depressive disorder were involved. Throughout the study, researchers collected more than 125,000 photos. The app was able to identify early symptoms of depression with a 75% success rate. Participants took these photos while answering a key question from the Patient Health Questionnaire PHQ-8 depression survey, which provided a context for image capture.

This research, detailed in a paper on the arXiv preprint database, will be presented at the Association of Computing Machinery’s CHI 2024 conference. The findings highlight a promising digital mental health technologies future, potentially making MoodCapture publicly available within the next five years.

MoodCapture Technology

The technology behind MoodCapture represents a significant advancement in digital mental health. It uses a process similar to facial recognition software but tailored to detect mood changes indicative of depression. By analyzing over 125,000 images taken passively through a user’s phone camera, the app’s AI model correlates specific facial expressions and environmental factors with the onset of depression without any active input required from the user and achieves a 75% accuracy rate. By fine-tuning this model with patient input, researchers hope to improve accuracy to 90%

Implications

This passive, non-intrusive approach allows real-time mood analysis and suggests therapeutic interventions based on detected symptoms. Researchers emphasize the app’s potential in providing immediate support and bridging the gap between the need for mental health intervention and access to care. 

The study showcases the feasibility of such technology and points towards future enhancements, including privacy measures and personalized diagnosis, paving the way for a new era in mental health care.

Reference

Nepal, Subigya, Arvind Pillai, Weichen Wang, Tess Griffin, Amanda C. Collins, Michael Heinz, Damien Lekkas, et al. “MoodCapture: Depression Detection Using In-the-Wild Smartphone Images,” February 25, 2024. https://doi.org/10.1145/3613904.3642680. https://arxiv.org/abs/2402.16182