DeepMind CEO Demis Hassabis at a 2017 event in China. Image from Alphabet
Even worse than hospitalization for a disease or injury, inpatients in medical institutions may also suffer from secondary diseases that may pose a serious threat to their health. According to the Centers for Disease Control and Prevention, acute kidney disease kills 500,000 American patients every year. Acute kidney injuries can be fatal, and they are a real problem for doctors. They are not only difficult to detect, but can also lead to rapid deterioration of patients.
However, through artificial intelligence systems, DeepMind's solutions can help doctors detect potential kidney damage 48 hours in advance, which can solve the problem in this precious time and prevent 30% of patients from getting worse in advance.
According to a paper published in Nature, DeepMind outlines its work with the U.S. Department of Veterans Affairs, which uses anonymous data to develop machine learning tools to correctly predict that nine out of 10 patients will need dialysis later. The study was carried out by DeepMind, University College London, University College London Hospital, and the Department of Veterans Affairs (VA) of the United States.
To address this problem, DeepMind worked with the U.S. Department of Veterans Affairs. DeepMind trained an in-depth learning system using data from more than 700,000 patients in the US Veterans Affairs Medical System. With this system, 55.8% of acute kidney injury can be predicted 48 hours before standard clinical diagnosis.
In the future, DeepMind hopes to combine this technology with Streams, a medical mobile assistant that can worsen patients'condition to doctors and enable communication between clinical teams and review of medical information so that they can make more effective treatment decisions.
DeepMind also announced Wednesday that researchers at University College London have successfully audited Streams and have been using them in local hospitals since 2017. They found that doctors could save two hours a day and allow them to see the most urgent 15 minutes.
"These findings form the cornerstone of our long-term vision of preventive health care and help doctors intervene in a proactive rather than passive way," the paper wrote. DeepMind's work in detecting kidney damage follows similar work done by the company last year, using AI to detect more than 50 eye diseases that threaten vision.