Scientists from the United States diagnosed depression at an early stage using Instagram photos. This site writes Takie dela.
Researchers from the University of Vermont, together with scientists from Harvard analyzed more than 40 thousands of photographs of users Instagram and developed an algorithm that can detect depression at an early stage in 70% cases.
There are compelling reasons for prioritizing research in the analysis. Instagram for health screening. Currently users Instagram publish almost 100 millions of new messages per day, and the frequency of new users Instagram lately ahead Twitter, YouTube, LinkedIn and even Facebook.
The results of the study were published In the magazine EPJ Data Science.
According to the World Health Organization, with depression, which is one of the most common mental disorders in the world, almost 300 millions of people live. But here it is possible to diagnose it correctly only in 42%.
The main tool for identifying symptoms of the disorder are questionnaires with which the patient assesses the intensity of his depression. Specialists rely on these data when they track the dynamics of the patient's condition.
But alas, this method does not reveal the full picture, as people often hide some manifestations or do not attach importance to them.
Psychiatrists have long sought additional effective methods for assessing the patient's condition, so the method of self-assessment alone is not enough. One of the directions was the development in the field of analyzing large amounts of data.
In early August, a team of scientists from Harvard and the University of Vermont presented a new technology based on the results of the questionnaire CES-D (Center for Epidemiologic Studies Depression Scale).
Of the thousands of photos analyzed almost 44 at the first stage of the study, it turned out that the participant had previously been diagnosed with a 71. At the second stage, participants were asked to take a test to determine the intensity of diagnosis.
During the survey, each patient was asked to rate how much he agrees with statements like "I often feel depressed" or "I wanted to hurt myself."
At the next stage, scientists figured out the ratio of the total number of photos per day, likes and comments under the photos.
Then, using the facial recognition system, scientists determined how often people appear in patient photographs. In addition, experts analyzed the color model of images and the most frequently used filters.
The result showed that those people who are prone to depression use blue and gray shades much more often. And the overall color of the photo turned out to be darker.
The most popular image processing filter in this group has become Inkwellmaking the image black and white. At the same time, study participants who did not have symptoms of depression often chose a “warm” filter. Valencia.
For those volunteers who are prone to depression, self-image most often appeared, and there are far fewer collective photos in their profiles than those who are not prone to depressive states.
Another hypothesis, which was expressed by scientists, is that photographs published before the first clinical diagnosis and after the start of treatment are significantly different.
Ratings of people in posts Instagram According to the general semantic categories can distinguish the messages made by depressive and healthy people.
Scientists have explained this ratio by the reduction of social contacts in people with mental disorders. According to the authors of the technology, in the future such methods will help to significantly reduce the number of false diagnoses in the world psychiatry.