DeepMind develops AI tool to predict kidney disease in advance

1 August 2019 (Last Updated December 23rd, 2019 10:22)

Google division DeepMind has partnered with the US Department of Veterans Affairs (VA) to develop an artificial intelligence (AI) tool that could predict acute kidney injury (AKI) up to 48 hours in advance.

DeepMind develops AI tool to predict kidney disease in advance
DeepMind trained AI technology using a dataset that comprised electronic health records of 703,782 adult patients. Credit: Gerd Altmann from Pixabay.

Google division DeepMind has partnered with the US Department of Veterans Affairs (VA) to develop an artificial intelligence (AI) tool that could predict acute kidney injury (AKI) up to 48 hours in advance.

AKI, which occurs when a person’s kidney functions abnormally, is associated with quick deterioration. According to DeepMind, up to 30% of cases could be addressed with earlier intervention.

The company applied AI technology using a dataset that comprised de-identified electronic health records of 703,782 adult patients from 172 inpatient and 1,062 outpatient VA sites.

Compared to existing techniques, the new technology was able to forecast the kidney disease up to 48 hours earlier.

DeepMind noted that the tool correctly predicted nine out of ten patients whose condition deteriorated to the dialysis stage.

According to the partners’ publication in the Nature journal, the AI model predicted 55.8% of all inpatient AKI episodes and 90.2% of all AKI cases that needed subsequent dialysis.

It is expected that these predictions could help in earlier preventative treatment and eliminate more invasive procedures such as dialysis.

The model has also been designed for future application in additional causes of diseases and deterioration such as sepsis, added the company.

DeepMind also unveiled results from research conducted by the University College London of its mobile medical assistant for clinicians called Streams.

The Royal Free London NHS Foundation Trust has been using the app since early 2017. Streams leverages the current national AKI algorithm in the UK to notify on patient deterioration.

The app also helps in the review of medical information at the bedside and facilitates instant communication between clinical teams. At the NHS Trust, Streams has been reported to save up to two hours per day for clinicians.

The latest research showed that the app could help review urgent cases within 15 minutes or less, with fewer AKI cases being missed.

While the app does not use AI currently, the company intends to integrate with predictive models.

In November last year, Google announced plans to absorb DeepMind’s health technology unit, including the Streams app team, into its newly formed health subsidiary, Google Health.