DeepMind gives AI new skills: Predicting patient's condition

After conquering Weiqi, one of DeepMind's biggest goals today is to be in the medical field. The problem that DeepMind recently tried to solve was the worsening of the disease that often occurred in hospitals. To this end, DeepMind provided his AI system with a history medical record of 70多万名 American veterans, hoping it would learn to predict which symptoms could lead to death without treatment.

This project is a collaboration between DeepMind and the United States Veterans Administration (VA). The aim is to get the top researchers in the AI field to work with top U.S. government doctors and medical experts to tackle the problems facing the medical world. It is worth mentioning that, for privacy protection, the personal information in all veterans medical data obtained by DeepMind has been removed.

In simple terms, the problem of the patient's worsening condition is human error. The study represents the US government's desire to use artificial intelligence to reduce human error in important areas. After all, AI is bringing a subversive revolution to the medical world, and this research will be an important cornerstone of the revolution.

In the traditional patient management process, one of the duties of a nurse is to monitor the patient's condition. However, because we can not provide each patient with a nurse for direct supervision, the majority of patients are in the condition of ECG, breathing apparatus, such as electronic devices for remote monitoring. Nurses and doctors take turns to make rounds and spend a little time on each patient; The central nurses ' station will also be responsible for overseeing all alarms. But the truth is, most patients are not being watched.

If DeepMind can make AI succeed in learning "why the symptoms of the patient worsen", in theory, this means that the machine will be able to oversee the patient. The machine can also take care of the patients 24 hours uninterrupted. While this does not necessarily immediately solve the problem of medical human error (the third cause of death in the United States), it may effectively reduce its incidence.

DeepMind also chooses to solve a specific problem, which will lay a solid foundation for overcoming the problems of the future. DeepMind said on the blog:

We will focus on acute renal injury (AKI), one of the most common conditions prone to deterioration, and an area where both DeepMind and VA have expert experience. This is a complex challenge because predicting AKI is not easy. This is not only because of the sudden onset of AKI and often without symptoms, but also because the risk factors associated with it are very common in hospitals. AKI can occur in people of any age and often after a routine operation such as hip arthroplasty.

As one of the world's top AI research institutions, DeepMind's shift from a "game" to a potentially lifesaving study will greatly improve the level of medical AI. With the wave of aging around the world, the problem of human error in the hospital will not be slow to solve. Excerpt from: http://www.sohu.com/a/224601385_354973 reference: https://thenextweb.com/artificial-intelligence/2018/02/26/ googles-deepmind-teaches-ai-to-predict-death/