Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Researchers design wearable tech that can sense glucose levels more accurately

uwaterloo.ca

Post reply on HN
Print view View original

41–50 of 171 posts

Re: Researchers design wearable tech that can sense glucose levels more accurately

#41

Post by caseyy »

blackeyeblitzar wrote: post #32
IndeterminateR wrote: post #28

Earlier quoted context omitted.

CGMs are probably going to still be more accurate, so the watch may not be ideal for Type 1, but could be nice for people with pre-diabetes who are trying to optimize their health. I'm also curious how accurate they will be.

CGMs themselves are still inaccurate compared to needles. And even needle based meters have a lot of different levels of accuracy. A CGM, even once calibrated, may be off by as much as 10%.

The value is in clinical application though. For closed loop systems, making a clinical decision for insulin dosing every 5 minutes is life-changing, even if the dosage is 10% off. I’d say even with compression lows, most systems are self-balancing enough to produce better results than a self-monitoring person could.

Besides, 10% off doesn’t often matter:

At 0-70 mg/dl the pump should suspend insulin either way. At 110-600 mg/dl the pump should ensure enough IoB by bolus, increase basal, and monitor either way. In that 70-110mg/dl the 10% MARD kind of matters for clinical decision-making, but not much. 90mg/dl is about as healthy as 81 and 99.

Patients are sometimes fussy by this inaccuracy but forget the tremendous benefit of trend indicators, let alone closed loop systems. Both of these have a much much larger positive impact to health than blood glucose being 10% above or below target impacts health negatively.

CGM with <=10% MARD, whether in wrist form factor or sensor form, is good enough for treatment. Of course, same as most readers, I have my doubts about this article.

Re: Researchers design wearable tech that can sense glucose levels more accurately

#42

Post by azinman2 »

hn_throwaway_99 wrote: post #29
sheepscreek wrote: post #26

Earlier quoted context omitted.

> they didn't do any testing, at all, with actual humans (or animals) They did mention in the article that clinical trials are on-going.

Correct, but I'm extremely skeptical, and that sentence had my bullshit detector alarms screaming even louder: > “We have a minimum viable product that’s already being used in clinical trials, and while there’s more work to be done, we’re much closer to a full marketable device,” Shaker said. Absolutely no information about what this "clinical trial" entails, or what phase it was. Most importantly, to get an initial…

But they have a minimal viable product! It’s viable! /s

Re: Researchers design wearable tech that can sense glucose levels more accurately

#43

Post by azinman2 »

> “We have a minimum viable product that’s already being used in clinical trials, and while there’s more work to be done, we’re much closer to a full marketable device,” Shaker said.

I see this a lot. People seem to ignore the “viable” part of MVP. If there’s more work to be done to make it a full marketable device, it isn’t viable in its current stage.

I wish them luck.

Re: Researchers design wearable tech that can sense glucose levels more accurately

#45

Post by mmooss »

jimrandomh wrote: post #27

This is yet another in a long line of glucose-measurement devices designed to sell to unsophisticated research grantmaking agencies, rather than to diabetics. Making a device that "measures" blood sugar in a watch form factor is easy, and many research groups have done so. Making one that's accurate enough to compete with the CGMs that are already on the market is a different matter entirely.

What about this research indicates to you that it doesn't address the accuracy issue?

Re: Researchers design wearable tech that can sense glucose levels more accurately

#46

Post by Hard_Space »

kumarvvr wrote: post #15

How can an algorithm be "artificial intelligence algorithm"? Does it mean AI models are used for data fitting? Or clustering? For data generated in such low scales, wouldnt statistical methods or procedural methods be sufficient or efficient or both?

If you hard-code effective learned distributions from a trained model, I suppose that could be described as an 'AI algorithm', even though the final output is a flat algorithm.

Re: Researchers design wearable tech that can sense glucose levels more accurately

#47

Post by caseyy »

azinman2 wrote: post #42
hn_throwaway_99 wrote: post #29

Earlier quoted context omitted.

Correct, but I'm extremely skeptical, and that sentence had my bullshit detector alarms screaming even louder: > “We have a minimum viable product that’s already being used in clinical trials, and while there’s more work to be done, we’re much closer to a full marketable device,” Shaker said. Absolutely no information about what this "clinical trial" entails, or what phase it was. Most importantly, to get an initial…

But they have a minimal viable product! It’s viable! /s

Well… yes. Looking at the article, it satisfies the MVP requirements for getting grant funding. A bar GlucoWatch cleared more than two decades ago, though it could never quite clear the bar of clinical viability.

Theranos also had an MVP in this sense :)

Re: Researchers design wearable tech that can sense glucose levels more accurately

#49

Post by winter_blue »

vlovich123 wrote: post #48

> and microcontrollers, which process the radar signals using artificial intelligence algorithms So glad we've rebranded primitive ML and basic control mechanisms as AI.

Or maybe just a signal processor is being rebranded as AI.

Soon a Fast Fourier Transform will be rebranded as AI as well.

Re: Researchers design wearable tech that can sense glucose levels more accurately

#50

Post by caseyy »

This all sounds exciting, and good luck to all researchers earnestly working on it. If RF blood content diagnostics become possible, blood glucose sensing would be just the start.

But as for this guy and his invention, don’t forget that in clinical setting, Theranos has shown more evidence of their product working. And GlucoWatch, a similar idea two decades ago, was FDA-approved and made it to market though still wasn’t clinically useful due to poor accuracy. Then, I’m not even talking about the charlatan cottage industry around glucose sensing watches, nor am I talking about how CGMs are a (generally) solved problem in diabetes.

Let’s wait for some clinical trials of the applied blood glucose sensing before we pop the champagne? It quite likely won’t happen, welcome as that invention would be.

Post reply on HN