UK-based Relative Health Limited has developed a novel AI approach to obtaining Blood Pressure from consumer health tech.
The system takes in synchronous electrical and optical sensor data acquired from the wrist or arm and utilises a novel Machine Learning approach to derive continuous Blood Pressure. The team have achieved a British and Irish Hypertension Society validation score for accuracy of “B/A” for Systolic/Diastolic Blood Pressures respectively, high enough for product certification. The system was trained on a 26,000 record dataset and has been ported to run on Huawei’s SmartWatch2 as the first operational deployment.
CTO Chris Crockford said of the development:
“Our work demonstrates a significant leap forward towards the ability to monitor Blood Pressure continuously without the need for uncomfortable inflatable cuff systems. With the current moves to standardize the certification of Blood Pressure devices this comes at a perfect time. As we migrate to a larger GPU cluster we are confident of reaching the desired gold standard A/A certification providing our hardware partners a step change in certified hypertension management.”
The latest UK National Institute of Health Research clinical trial is under way within the NHS utilising the AndroidWear Huawei SmartWatch2. The team are now working to optimise the performance for different devices, dealing with for the nuances of data sampling rates and physical locations of sensors for different manufacturer’s devices.
The work has increased the IP portfolio of the company considerably and was part funded by InnovateUK.
Relative Health has the aim of becoming the number one provider of certified hypertension algorithms for consumer devices, driving change for both the consumer and the medical world.