Google’s A.I.-centric DeepMind division has announced a new partnership with the U.S. Department of Veterans Affairs (VA) aimed at predicting kidney problems earlier. Specifically, the focus is on Acute Kidney Injury (AKI), which is one of the most commonly occurring patient deterioration issues the department sees. The problems with AKI stem from the fact that, as well as being common, the condition sometimes shows no visible symptoms and is often fatal by the time it’s noticed. In fact, as many as 11-percent of all in-hospital deaths are the result of patient deterioration not being caught early enough. That’s because Patient deterioration can lead to widespread, difficult to control infections and other complications. DeepMind and the VA are hoping to use A.I. to analyze patient data and find a way to predict when AKI might become a problem, which could also lead to a better understanding of other deterioration issues.
In total, DeepMind and the VA will analyze around 700,000 historical, depersonalized medical records as part of the project. Machine learning will be brought to task to see if there is any useful way for the technology to assist in accurately and quickly identifying both risk factors and when deterioration began. More directly, the goal is to find ways to improve the algorithms currently being used to detect AKI, giving doctors and nurses more time to intervene. Predicting both of those metrics accurately may give veterans served by the VA a better chance of survival when facing these types of medical problems. However, perhaps more importantly, it may pave the way for A.I. in this type of circumstance more generally. Success could, with consideration for the fact that AKI and similar conditions can affect people of any age and often occur after other medical procedures, ultimately save millions of lives.
It’s important to bear in mind that the current iteration of this research is experimental at best. So it could also result in no improvement at all. That’s not necessarily the most likely outcome and, if that happens, it doesn’t mean that A.I. will never have a use in this particular part of the medical profession. To begin with, the technology has already shown promise in several other medical applications. It may just be that it hasn’t progressed far enough at this juncture to have an impact here.