Home Medizin Smartphone-App nutzt KI und Gesichtsbildverarbeitungssoftware, um den Beginn einer Depression zuverlässig zu erkennen

Smartphone-App nutzt KI und Gesichtsbildverarbeitungssoftware, um den Beginn einer Depression zuverlässig zu erkennen

von NFI Redaktion

Researchers from Dartmouth have developed the first smartphone application that utilizes artificial intelligence paired with facial image processing software to reliably detect the onset of depression before the user even realizes something is wrong. The app, named MoodCapture, uses the front camera of a phone to capture a person’s facial expressions and surroundings through regular use, then analyzes the images for clinical indications related to depression. In a study of 177 individuals diagnosed with severe depressive disorder, the app identified early signs of depression with 75% accuracy. These findings suggest that the technology could be publicly available within the next five years as it continues to be developed, according to researchers from the Department of Computer Science in Dartmouth and the Geisel School of Medicine.

The team published their article on the Preprint Database arXiv before presenting it at the CHI 2024 conference of the Association of Computing Machinery in May. Papers presented at CHI undergo peer review prior to acceptance and are published in the conference proceedings.

This is the first instance of natural images being used „in the wild“ to predict depression. There is a movement towards digital technology in mental health to ultimately develop a tool that can predict the mood of individuals diagnosed with severe depression reliably and non-invasively.

According to Andrew Campbell, corresponding author of the article and Albert Bradley 1915 Professor of Computer Science at Dartmouth, „People use facial recognition software to unlock their phones hundreds of times a day. MoodCapture utilizes a similar technology pipeline of facial recognition technology with deep learning and AI hardware, hence there is enormous potential to expand this technology without requiring additional input or burden for the user.“

For the study, the application captured 125,000 images of participants over a 90-day period. Participants agreed to have their photos taken with the front camera of their phones but were unaware of when this occurred. The first group of participants was used to program MoodCapture to detect depression, with their facial expressions captured while answering the question, „I have felt down, depressed, or hopeless.“ The question is from the Patient Health Questionnaire-8 (PHQ-8) used by doctors to detect and monitor severe depression.

The researchers used image analysis AI for these photos, allowing MoodCapture’s prediction model to learn to correlate self-reports of feeling depressed with specific facial expressions – such as gaze, eye movement, head positioning, and muscle stiffness – and environmental features like predominant colors, lighting, photo locations, and the number of people in the image.

The concept is that MoodCapture analyzes an image sequence in real time every time a user unlocks their phone. The AI model establishes connections between facial expressions and background details crucial for predicting the severity of depression, such as momentary changes in facial expression and the person’s environment. Over time, MoodCapture identifies user-specific image features. For example, if someone consistently appears with an expressionless face in a dimly lit room over an extended period, the AI model may infer that the person is experiencing onset depression.

The researchers tested the prediction model by having a separate group of participants answer the same PHQ-8 question while MoodCapture photographed them, analyzing their photos for signs of depression based on data collected from the first group. The MoodCapture AI correctly identified whether this second group was depressive or not with 75% accuracy.

Nicholas Jacobson, co-author of the study and Assistant Professor of Biomedical Data Science and Psychiatry at the Center for Technology and Behavioral Health in Dartmouth, noted that technologies like MoodCapture could help bridge the significant gap between the onset of depression symptoms and access to mental health resources. Traditionally, assessments overlook the fluctuations in depressive symptoms experienced by individuals over shorter time frames, which can exacerbate the impact of depression.

The study, led by Jacobson and funded by the National Institutes of Mental Health, leverages deep learning and passive data collection to detect depression symptoms in real time. It builds on a study from Campbell’s lab in 2012 that collected passive and automatic data from participants‘ phones in Dartmouth to assess their mental health. The advancement of smartphone cameras since then has made it possible to collect the kind of „passive“ photos taken during normal phone use, according to Campbell.

This work shows that passive photos are key to successful mobile therapy tools, as they capture mood more accurately and frequently than user-generated photos or selfies without requiring active engagement. The next steps for MoodCapture include training the AI on a broader diversity of participants, improving diagnostic capabilities, and enhancing data privacy measures. The researchers envision an iteration where photos never leave a person’s device but instead are processed on the device to extract facial expressions associated with depression and convert them into code for the AI model.

In the meantime, the application’s consumer-side accuracy could be enhanced by designing the AI to expand its knowledge based on the facial expressions of the individual using it. This way, the model wouldn’t have to start from scratch but could build upon the user-specific expressions for improved accuracy.

Overall, technologies like MoodCapture have the potential to revolutionize mental health by providing real-time support without burdening the healthcare system, ultimately aiming to interrupt and treat depression before it worsens. By focusing on the moment and capturing rapid changes in depressive symptoms, these tools can play a crucial role in early intervention and personalized treatment of depression.

Related Posts

Adblock Detected

Please support us by disabling your AdBlocker extension from your browsers for our website.