DeepMind is joining forces with the U.S. Department of Veterans Affairs to develop artificial intelligence technology that can help doctors predict when the condition of their patients may suddenly begin to deteriorate.

The project, which was announced on Thursday, will initially focus on facilitating faster detection of acute kidney injury. This is one of the conditions most commonly associated with sudden patient deterioration and often doesn’t show any obvious symptoms. Moreover, it can strike people of all age groups, a fact that makes understanding what’s wrong particularly difficult.

DeepMind hopes AI can address the challenge. As part of the new collaboration with the VA, the Alphabet Inc. subsidiary will analyze about 700,000 anonymized medical records for patterns that could potentially be used to identify cases of acute kidney injury. The hope is to find common indicators of the condition that an AI model can check to perform diagnosis.

Such pattern-based analysis is showing promising results in other medical fields. Verily Life Sciences LLC, another Alphabet subsidiary, earlier this week revealed that it has developed a method of analyzing blood vessels in the back of the eye to assess a person’s susceptibility to heart disease.

DeepMind’s new project could potentially make an even bigger impact if it proves successful. In the blog post announcing the collaboration with the VA, DeepMind Clinical Lead Dominic King wrote that late or incorrect response to patient deterioration is one of the biggest problems that hospitals face today.

King also stressed that the 700,000 medical records used in the project will be anonymized before they’re handed over to DeepMind. That’s significant because the group was found to have received “inappropriate” access to patient data as part of an earlier project that likewise focused on tackling acute kidney injury. Since then, DeepMind has formed an in-house ethics unit to avoid such situations.

The company has also been busy with other projects in the healthcare sector and beyond. In January, DeepMind open-sourced a tool that enables AI researchers to apply principles from the field of psychology to streamline their projects.

The software was released shortly after the group published a paper detailing how its AlphaGo system had mastered chess and a similar game called shogi using the same algorithm. This is significant for several reasons, chief of which is that the AI demonstrated a level of adaptability beyond what conventional models possess.

Image: DeepMind