Investing in bespoke processes can sidestep the inefficiencies of ‘lazy artificial intelligence (AI)’ in medical device development, a CEO has said.
While developing its Microwearable Hydration Sensor, WearOptimo built its own proprietary AI/machine learning model upon realising that the allure of big data, pulled from myriad unstructured datasets, or ‘lazy AI’, as WearOptimo’s CEO, Mark Kendall refers to it, would be insufficient in meeting the company’s needs.
Go deeper with GlobalData
Discover B2B Marketing That Performs
Combine business intelligence and editorial excellence to reach engaged professionals across 36 leading media platforms.
“In the context of hydration, which is our first area of application, we’ve created a completely new digital biomarker for hydration, currently the only one,” Kendall tells Medical Device Network.
WearOptimo’s sensor platform accesses a thin layer within the skin to quantify an individual’s hydration levels.
“What that means is that we don’t have a huge volume of data points. We’re not ripping data from unsupervised sources. These are sparser data sets, but high quality and supervised,” Kendall explains.
To evaluate its platform, WearOptimo recently concluded its first clinical study, with a publication of the results forthcoming.
US Tariffs are shifting - will you react or anticipate?
Don’t let policy changes catch you off guard. Stay proactive with real-time data and expert analysis.
By GlobalDataThe study was in exercise induced dehydration, where subjects were placed in an environmental chamber set at 40 degrees, undertaking exercise that induced dehydration before replenishing fluids and doing so again for around six hours.
“In the study, participants’ hydration level would be inferred by the so-called gold standards. Now I say so-called because they’re not that great,” Kendall said.
“Blood draws, for instance, just tell you how salty your blood is. In the context of hydration, blood is a poor metric since dehydration does not happen uniformly through the body; the body has priority areas that it tries to maintain as long as possible and sacrifice other areas to maintain that. And blood is top of the tree,” Kendall explained.
Throughout the course of the study, WearOptimo’s sensors pulled out datapoints that were tagged with a certain percentage of dehydration against mass loss to establish a labelled dataset.
In applying AI/ML on untrained datasets, WearOptimo found that hydration levels could be determined with a receiver operated characteristic of 0.72. Kendall explained that this score is significant in contrast to blood draws, in which the best possible score is usually around 0.6.
“We outperformed gold standard blood draws in our very first clinical study. And there’s no way we’d be able to do this if we didn’t take this approach,” he said.
Looking ahead, WearOptimo anticipates that its sensor will be commercially available within the next two years.
“The reason why we can be so streamlined is that we actually do not need regulatory approval applications for hydration for our first markets,” Kendall continued.
“Rather than the traditional medtech approach for regulatory approval, this allows us to gain access to data more quickly, and with more fidelity, which in turn accelerates our clinical application.”
In closing, Kendall highlights that hydration is currently a largely untapped market, one the company believes it is one of the first entrants within with the aim to ideate on the current gold standard to quantifying hydration levels.
Kendall concluded: “We believe that we can become essentially the scale of Dexcom just doing hydration alone.”
