It is now possible to identify kidney injuries up to two days even before it takes place, thanks to a new AI tool. In the last few years, tech comapnies have developed several AI tools to track and maintain medical records. Further, the basic idea is to generate advance warning enabling doctors to intervene before its late. Artificial intelligence (AI) coached on health records can now detect acute kidney injury 48 hours prior to occurrence.
Of late, doctors regard AI as an aid for diagnosis, especially for childhood conditions and specific types of cancers. Unfortunately, very few tools undergo meticulous clinical trials. Thus, it is quite early to predict the success of AI.
AI Uses Creatinine Level to Detect the Probability of AKI
Nenad Tomašev from DeepMind Technologies and his co-workers trained in algorithms for this tool. Further, they have written this algorithm that hints at possibilities of AKI for an in-patient.
The team from DeepMind Technologies coached the AI using de-identified electronic health records of 703,782 veterans. Further, they were in-patients between 18 and 90 years during October 2011 – September 2015.
AKI leads to a drastic drop in the kidneys’ rate of filtering blood. As a result, it deceases urine production and waste products accumulate in the blood. For instance, creatinine is a by-product of muscle breakdown, and is a measure for AKI diagnosis.
AI analyzes creatinine levels from patient records and points out if the patient is at the risk of AKI. Researchers confirmed this after testing the patients for AKI using diagnostic tests.