In the future, the fate of coma patients will be decided by computer algorithms

Researchers in China will use an algorithm to determine whether a comatose patient wakes up again.

When a person falls into a coma, they lose all motor functions. Brain activity slowed down significantly. In most cases, no external stimuli, such as light or motion, can awaken these patients. It is difficult to determine their status in the future - will they wake up again?

According to Futurism, Chinese neuroscientists at the PLA Public Science and Hospital Institute in Beijing are working hard to develop a tool to help doctors predict exactly that. But they have a technological advantage that many generations of doctors did not have: machine learning . Such algorithms are part of a growing array of data analysis tools that can help family members and doctors make difficult decisions about the treatment of a patient, or help determine when is the right time to say goodbye.

This algorithm understands the recoverability of a specific patient.

The researchers gave machine learning algorithms to study fMRI data (functional magnetic resonance imaging) from thousands of coma patients. As a result, this algorithm understands the recoverability of a specific patient.

The results are really promising.

"We have successfully predicted that some patients who have regained consciousness after initial diagnosis are no longer hopeful of recovery," the researchers said.

The researchers found that the algorithm was 90% accurate , and they used the technique on more than 300,000 patients across China. They hope the technology could help more than the estimated 50,000 "patients with chronic consciousness disorders" in China.

Fate betting on machines seems to be a risky option, but coma patients are actually the ideal subjects to apply this type of machine learning technology - according to neurologist Pascal Kaufmann, neurologist. and founder of Starmind, a Swiss-based company that is developing artificial intelligence to help large company employees communicate with each other. In fact, machines are capable of analyzing this complex biological data much better than humans."Those machines do nothing other than what people are doing. They also look at the same data - they do the same. However, they do faster and more reliable than we fold. a million times ".

Researchers in Beijing do not think that machines should be made the final decision about the life or death of coma patients."When we notify the patient's family about the diagnosis of AI, we always tell them that their decision should be based on them from 20 to 50% only" - Yang Yi, a Department physician Neurosurgery at the PLA Public Hospital and researcher in the project said.

The Kaufmann agrees - the evaluation of a computer system is only of concern if it determines that a comatose patient has the potential to recover after the human doctors judge that there is no hope, not opposite."When the human doctor says the patient will never wake up again, it is a heartbreaking sight. The fact that you let a patient go as a result of the machine - that's something that shouldn't happen. " - Kaufmann said -" I think you should only pay attention to the results of the machine if someone tells you that there is still hope. "

Now, when we have technology that can better predict whether coma patients will regain consciousness, Kaufmann said it would be dangerous for human doctors to analyze these data alone. It's like a self-driving car - human drivers are much more prone to accidents than drivers entrusted with self-driving cars."For the right doctor to decide whether a person will wake up or not will be very dangerous because the doctor's error rate is much higher than the machine" - Kaufmann said.

Currently, this algorithm is only applicable to coma patients.

At this point, the above algorithm is only applicable to coma patients. And that is probably a good thing. Machines are actually better than doctors in assessing a patient's condition, Kaufmann said. But they do not have the soft skills that patients look for doctors to treat them to have."The problem is, when it comes to human interaction [with patients who are not in a coma], of course, the human doctor is superior to the machine, because you can evaluate behavior, how they talk . - there are many factors that machines can't handle well ".

Allowing a computer to influence a decision regarding a patient's life or death is like a episode of Black Mirror, but that's probably a good thing. It is possible that machine learning algorithms will enter many more health care areas - they can analyze medical data collected from users through high-tech devices, or help surgery robots. The work done on patients is no longer (or very little) interoperability. With more data, they will become more and more accurate.

But what about a future where machines alone make the decision to unplug the life-sustaining system for coma patients? Probably a long time.