Thursday, November 2, 2017

Tech Briefing: AI Detects Suicidal Tendencies in People Using Brain Scans

In 2014, about 42,773 people in the United States committed suicides, according to the National Center for Health Statistics (NCHS), and suicide is the 3rd leading cause of death for 15 to 24 -year-olds and 2nd for 24-35-year-olds. The annual U.S. suicide rate has increased 24% over the past 18 years and keeps growing.

A terrifying facts, right? But what if we can identify individuals with suicidal thoughts before they commit suicide, and apply appropriate measures on them? I believe many people will be saved back and economic losses will be reduced.

In this week, researchers from Carnegie Mellon University conducted experiments by using artificial intelligence to identify individuals who suffer from suicidal thoughts.

Prior to the experiments, the researchers gathered 34 participants, half of whom were somewhat experiencing suicidal thought. In first experiment, all participants underwent functional magnetic resonance imaging (fMRI), and at the same time they were shown a bunch of words related to suicide and negative emotions such as “death” and “distressed “, as well as words relative to positive emotions such as “glory”. After analyzing results from participants who had already been identified as suicidal, the researchers could identify five regions in the brain and six words, facts that helped researchers identify suicidal patients. Using that information, researchers trained an algorithm to find out patients with potential suicidal thoughts. That algorithm ended up with correctly identifying 15 of the 17 suicidal patients and 16 of 17 members of the control group.

In second experiment, researchers also separated participants with suicidal thoughts into two groups: one that had attempted suicide, but another one that had not. The participants were guided by the same process, and a new algorithm was created by researchers based on the results. This time, the algorithm became more accurate than the prior one, identifying 16 of those 17 patients.

Since psychiatric illnesses are so complex that scientists are often confused by those illnesses’ uncertain active areas in the brain, some latent disorders such as depression often cause severe outcomes such as suicide. Now this artificial intelligence algorithm could assist scientists in more accurate diagnosis and more effective treatment than before.

Question:
Could you foresee any inhibitions or limitations on this AI algorithm? Such as technological or legitimate factors or else.



3 comments:

  1. Hi Yiming,
    I really like your post regarding the use of AI to identify those with suicidal thoughts. Suicide has been a very important issue with the rising rate in past few years as you mentioned. Based on the recent trends in the tech world, AI has become a very prominent project for some of the biggest companies.
    I personally like that the researchers are also using AI to implement it in the healthcare industry. However, I would be concern about its legitimacy because it will take time for the algorithms to truly understand the pattern or the symptoms.
    However, it won't be long till we see AI being used in most industries and with the wide-amount of use it will also become easier to implement and accuracy will increase too.

    ReplyDelete
  2. I find this post to be quite interesting. After hearing about so many tragedies, it is clear that there needs to be development of technology to further find ways to prevent suicide. I think that AI is amazing in that it can detect certain patterns, however, it would be interesting to see how effective it is based on range of age. Not only that, but I think that there will be restrictions based on who is able to take these tests. Will parents or supervisors be able to recommend someone to take it or will there be restrictions on that as well?

    ReplyDelete
  3. I think this is a great and very interesting article. Mental health research and awareness varies greatly from state to state in this country. I find the use of this particular algorithm very interesting and can see it as a potential tool to help prevent suicide.
    I think this tool could eventually be helped in other areas such as those individual who commit mass shootings/murders. I solemnly believe that mental records and exams should be utilized at the time of gun licensing and sales. Often times when these shootings occur these criminals have been labeled as unstable and often times nothing has been done to prevent their access to firearms. Mental illness has always been a huge issue and due to societal norms and issues has not been brought to light until recently.
    Health research in general should be looked at more closely and I definitely encourage studies such as this to be further funded and mandated.

    ReplyDelete