Suicide is the second leading cause of death among young people aged 15-34 in the United States, and clinicians can only use limited means to identify those who have suicidal tendencies. A paper published today in Natural Human Behavior documents a new machine learning technique that helps identify people who have suicidal thoughts. The researchers surveyed 34 young people, half of whom were suicidal and experimental. Each subject was examined by functional magnetic resonance imaging (fMRI) and gave three lists containing 10 words. All vocabulary is related to three aspects: suicide (such as "death", "pain" or "fatal"), positive influence ("carefree", "good", "naive") or negative influence ("boring" ", "evil", "guilty"). The researchers also used previously drawn neural signal maps that show the brain's emotional patterns, such as "shame" and "anger." Five brain localizations, along with six words, were found to be the best markers for distinguishing suicide patients from the control group. Using these locations and vocabulary, the researchers trained a machine learning classifier that correctly identified 15 suicidal participants in 17 suicides and 16 of 17 control subjects without suicidal tendencies. By. Subsequently, the researchers divided the suicidal patients into two groups, one with suicide experience (9), the other with no suicide experience (8), and trained a new classifier that correctly identified 17 16 patients in the patient. The results showed that mental health participants and those with suicidal thoughts responded significantly differently to words. For example, when a suicidal participant sees the word "death," the "shame" area in their brain is much brighter than the corresponding area of ​​the brain in the control group. Similarly, the word "trouble" will trigger more brain activity in the "sadness" area. This is the latest attempt to introduce artificial intelligence into psychiatry. Researchers are working on machine learning programs ranging from analyzing nuclear magnetic resonance spectroscopy to predicting major depression to identifying post-traumatic stress disorder (PTSD) from people's speech patterns. Earlier this year, Wired magazine reported on a number of researchers who built a system to identify people with suicide risk by analyzing health records, with an accuracy rate between 80% and 90%. Facebook is using text mining technology to identify users at risk of suicide or self-harm and then point it to mental health resources (see "Several issues with Facebook's suicide prevention tools"). Artificial intelligence has already set off waves in the medical field. Some algorithms are very good at detecting other problems in tumors and CT scans, and Jeffrey Newtown told The New Yorker that radiologists will eventually lose their jobs. He is one of the most important researchers in the field of deep learning. In fact, he said, "They should stop training radiologists now." In this case, the study is more likely to provoke new human-driven treatments than to let a doctor in the entire field lose his job. This paper points out that identifying different patterns and regions can open up new areas for brain stimulation technology. Determining specific emotional responses to suicide-related terminology can also be used by psychotherapists to treat patients. Shenzhen Ruidian Technology CO., Ltd , https://www.wisonens.com