Know Labs unveils first prototype of portable, noninvasive glucose monitor

On the heels of sharing study results demonstrating that its Body-Radio Frequency Identification, or Bio-RFID, technology could accurately measure blood sugar levels completely noninvasively, Know Labs has built the first iteration of the hardware that’ll house that technology.

In a Wednesday announcement unveiling the prototype, Steve Kent, Know Labs’ chief product officer, described the Generation 1 device as “the ultimate research and development tool.” It’s designed to fit in a user’s pocket or, when they need to perform an on-demand glucose test, in the palm of their hand.

According to Kent, it’s equipped with “more on-board computing power than a desktop PC, built-in machine learning capabilities, a long battery life, and it is fully configurable to support our many development initiatives.”

Developing the portable device took several years and “hundreds of iterations,” CEO Ron Erickson said in the release, and required the company and its partners to overcome “incredible engineering complexities.”

“The completion of Gen 1 marks a significant engineering achievement for the company and, more broadly, for innovation in medical diagnostics. Many have tried to noninvasively ascertain glucose and have not succeeded or remain years away from success,” Erickson said. “The Bio-RFID sensor is a novel technology that is leading the way to an entirely new branch of science, and Gen 1 takes us closer to our goal of enabling a better way of life for people living with diabetes.”

With the Gen 1 prototype in hand, Know Labs said in the announcement that it plans to spend the rest of this year continuing to validate the Bio-RFID technology. It’ll also continue to tweak the physical device, all with an aim of eventually submitting the system to the FDA for clearance—which would make it the first noninvasive glucose monitor authorized by the regulator.

The Bio-RFID sensor works by sending radio waves through the skin to measure molecular signatures in the blood, which Know Labs’ machine learning algorithms then use to compute the user’s blood sugar levels.

Know Labs is rapidly improving the technology’s accuracy in making those calculations. Results from two studies presented earlier this year showed that the sensor was measuring glucose levels with a mean absolute relative difference (MARD)—which indicates the variation between a sensor’s readings and a reference glucose measurement—of around 20%.

But in another study presented just a few weeks later, at the end of May, Know Labs described how adding a new machine learning model called a light gradient-boosting machine, or LightGBM, to the mix had narrowed the difference between its sensor’s readings and those of a Dexcom G6 continuous glucose monitor, for an overall MARD of just under 13%.

For reference, a glucose sensor with a MARD below 10% is generally considered to be highly accurate. Dexcom’s newly FDA-cleared G7 CGM boasts a MARD of just over 8% in children and adults, while Abbott recently clocked a 7.9% difference in a study of its own 14-day FreeStyle Libre 3 sensor.