Artificial intelligence will determine a depression photo on Instagram

The specialists at Harvard and Vermont universities taught artificial intelligence to identify the signs of depression photos in Instagram. But can social networks replace physicians in the diagnosis of mental disorders?

To teach the computer to recognize the symptoms of depression, the researchers collected for analysis 44 thousands of photos 166 users Instagram: 95 of them were healthy, and 71 diagnosed with clinical depression.

Experts take into account not only the characteristics of the image (tone, brightness, saturation, filters applied), but also other factors: the degree of user activity (number of publications in the day), the reaction of the followers (likes and comments) on the photos.

Associated symptoms:

Mood swingsDepression

The main visual signs of depression appeared to be images of low brightness and saturation of colors close to the blue end of the spectrum, a small number of likes, the frequent use of black and white filter, Inkwell. In healthy users of Instagram on the photos was more of the persons, the images were bright, and a favorite filter turned out to be Valencia.

The main visual signs of depression appeared to be images of low brightness and saturation of colors close to the blue end of the spectrum, a small number of likes, the frequent use of black and white filter, Inkwell.

After training the artificial intelligence was able to correctly identify the depression in 70% of cases, but often gave false positive results.

Earlier studies have shown that General practitioners are mistaken in the diagnosis of depression in 58% of cases, but successfully exclude it in 81% of healthy people. Thus, while the computer and the human to work better in tandem.

The authors write that in addition to depression, Instagram can help physicians in the early detection of other mental disorders: postpartum emotional disorders, alcoholism, suicidal tendencies, and PTSD(post traumatic stress disorder).

Valencia favorite filter healthy people.

Source: N 1

Associated diseases:

Mental disordersAlcoholism Post-traumatic stress disorder