Modern medicine aims to prevent and anticipate the development of serious life-threatening diseases. However, the strength and knowledge of a person is often not enough to predict certain complications. The AI DeepMind from Google, which is currently learning to predict the death of people, may come to the aid of doctors in the future.
As is often the case in a similar situation, already known clinical cases were used to train AI. Developers from Google “fed” DeepMind the data of nearly 220 thousand adult patients treated between 2009 and 2016. Patient information was provided by the University of California San Francisco Medical Center and the City of Chicago Medical University. In addition, the scientists used information received from the US Department of Veterans Affairs, which provided access to data on 700,000 monitored former military personnel. In total, more than 46 billion items were added to the database on a variety of health indicators. At the moment, the algorithm is aimed at finding conditions such as acute renal failure and pneumonia. These states were not chosen by chance, because, according to the authors of the work, “Acute renal failure or pneumonia can affect people of any age, and they often begin after routine procedures and surgeries. 11% of all deaths in the hospital are caused by a deterioration in the patient's well-being, and changes in the patient's condition cannot always be noticed at an early stage.
The system is now able to predict the probability of a patient dying within 24-48 hours, but the main goal of the project is to learn how to predict death in the next 12 months. In addition, it is planned to expand data on deadly pathological conditions. Such use of artificial intelligence for medical purposes can be used not only in order to be ready for resuscitation measures, but also to prevent the development of severe conditions in patients.