49 comments

[ 2.8 ms ] story [ 89.2 ms ] thread
Eh a lot of “these particles bonded with glucose - so now we can measure blood sugar levels” that’s a massive stretch from where they are at.

It’s cool, but nothing close to what’s described.

I don’t think the magnetic particles interact with the glucose, they’re just a magnetic transducer basically. The main measurement device is a hydrogel that reacts with the glucose and swells, and they measure how far the magnetic particles move using a magnetometer to sense the change in field strength and back out the concentration (and pH)

The linked Nature Comm paper seems pretty thorough, and they are reporting really high sensitivity (1000x what you’d need for typical blood glucose levels)

https://doi.org/10.1038/s41467-024-47073-2

Okay sure, I shortened it a bit. My point is the discovery wasn’t directly related to the claims. Your description is more apt than the hyperbole in use cases. To accomplish what they describe would need a lot more research
To be fair, they do call it a “proof of concept study” in the press release
I wonder about the reference for the measurement, or precision versus accuracy. Still super interesting application of hybrid common and specialized tech.
Joking aside, you can actually use a wearable non-invasive electromagnetic sensor device to accurately measure glucose level [1].

[1] Meet the inventor: Professor Adrian Porch from Cardiff University:

https://www.med-technews.com/medtech-insights/latest-medtech...

Type-1 diabetics have no working pancreas, and need something that meets the highest standards before it gets approved for use. Is this product up to those standards?
I think a MARD of <= 14 would be required to be useful.
10% is what you really need to have excellent control
I‘m not aware of any accurate noninvasive device on the market. Recently there‘s been an influx of fake watches that claim to do so, but they mostly make up the values. The linked article mentions that research is ongoing - can you point me where such a device is being sold?
I'm not sure if the device is currently being sold since they need to get the approval for any medical devices. For results and accuracy please check their papers from several years back [1], [2]. I think may be you are right, for transparency they need to release all the accuracy, sensitivity and specificity of the non-invasive glucose detection device, and from the publication title it seems only sensitivity results are currently being published. Personally I have one patent pending on non invasive ECG anomaly detection that cover both arrhythmia and ischemia with comprehensive 99% accuracy, sensitivity and specificity, and high accuracy of non invasive detection technique is certainly achievable.

[1] Microwave Noninvasive Blood Glucose Monitoring Sensor: Penetration Depth and Sensitivity Analysis (2018)

https://ieeexplore.ieee.org/document/8428850

[2] Microwave noninvasive blood glucose monitoring sensor: Human clinical trial results (2017):

https://ieeexplore.ieee.org/document/8058721

In order to bring a CGM to market, in many areas of the world you need to perform a number of clinical trials, and a certain MARD upper bound has to be demonstrated. Also, during development certain ISO/IEC standards must be followed, which cover continuous medical & information risk assessment and mitigation. It’s usually a process of up to a few years even after you have a strong belief that it performs well enough.
> accurately

Not nearly accurate enough to make insulin dosing decisions, which is all diabetics care about.

For the foreseeable future, non-invasive glucose monitors will be like the scales that measure your body fat through electrical impulses. A neat toy, but not actually useful for medical purposes.

> Not nearly accurate enough to make insulin dosing decisions, which is all diabetics care about

It might be all type-1 diabetics care about, but for the 90-95% of diabetics that are type-2 much less accuracy is needed.

Well, 25% of those type-2 diabetics take insulin, but yes, those that don't take insulin may be able to derive some value from a less accurate sensor.

I just don't know why you'd choose the less accurate thing when it's your health.

I think the non-invasive stuff makes the most sense for people who don't have diabetes yet. Nothing will be covered by insurance for them and the non-invasive tech might be cheaper, and it can nudge them to make healthier eating decisions.

This is enough for non-diabetics who are curious about their diet and looking for data on blood sugar through the day to adjust their diet.
Don't be too pessimistic regarding the accuracy of non-invasive screening technique, before you know it you will probably be using one of the solutions. This high accuracy digital diagnostic and biomarker problems are way much easier than the level 5 autonomous driving and AGI, and please don't let other people to tell you otherwise. Personally I think they (company/govt agency/research labs) should hire more reseachers and PhD in this field of bio medical engineering to accelerate the efforts for the benefit of humankind. Not surprisingly the medical doctors associations (I am looking at you American Heart Association - AHA) will be up in arms against automated diagnostics screening eventhough you have 99% accuracy or better because of their biased perceptions due to their job insecurities.
The fact that Apple has not gone for it yet tells a lot about the actual usability of such devices imho.
I have been playing around with various sensors on the skin, one that is interesting is IR absorption which can read glucose as well. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5772583/ I got some interesting results with combining sensors and using ML to identify different labels, even managed to identify left and write arms?!? I made a few notes here https://nix64bit.com/identification.html
> left and write arms

How do people make this mistake

I was surprised to read "Employing this strategy, the researchers measured glucose concentrations as small as a few millionths of a mole" in an article published on an NIST website.

I fear that they, like so many organizations, must employ non-experts to interview experts and write stories. But getting the unit of concentration wrong? (Later in the sentence they do note that moles refer to quantity, not concentrations, so the whole sentence is self-contradictory.)

This problem is so common I think it might make sense to suggest a common solution. Perhaps an organization specializing in X ought to hire one member of the writing team who has a degree in X, to look over material before it goes out to the world. Or at least pay one of the experts on a per-case basis, say a day of salary for a half hour of work, to go over material before it is published.

I honestly just hope it's satire, it almost feels on purpose
I honestly just hope it's satire, it almost feels on purpose
(comment deleted)
Technical correctness and public relations are natural enemies. The original material comes from scientists who know what they are doing, but they aren't allowed to write their own press releases. For scientific papers, the scientists are already encumbered by layers of review that make it difficult to meet deadlines for publishing.
I instinctively read it as referring to molar concentration. It is actually common in chemistry to say "10-mole solution", which means 10 moles per litre. The per litre part is implicit, which confused me greatly during undergrad classes. The actual unit of 1 mol/L is called a molar.
Yup, I knew a postdoc once who didn't like to say litre; the first time I first heard him say "mole per dm cubed" I didn't know what he meant. (He also had an accent.)
These sensors might work well in the lab, but people have been trying various spectroscopy based noninvasive sensing techniques for the better part of 15 years, with limited success.

The problem is not that you can get a nicely correlated signal when holding a stationary phone against a stationary well of known concentration, the problem is how does your sensor perform when:

1. You're exercising (sweat getting between the sensor and the skin)

2. You're on vasoconstrictors (smoking or blood pressure medications)

3. You're taking medication that "looks" similar to glucose (in electrochemical sensors that's ~~ibuprofen~~ acetaminophen, but the number of magnetic analogues, especially in the presence of interference, is surely greater)

4. You have low perfusion

5. You're sleeping on/compressing the sensor. This changes how the blood flows, and needs to be accounted for or you'll risk reading a false low value.

These are just a handful of the problems that modern CGMs face, and only a few of them have working solutions.

The article is not about a noninvasive sensing technique.
Magnetometers are noninvasive! They don't penetrate the skin
That's not how it is used here.

The sample liquid containing glucose, e.g. wine or possibly blood (the sample might need to be diluted first with pure water) is put on a gel that has volume changes depending on the amount of glucose.

The volume changes are detected magnetically because the gel is coated with a ferromagnetic material.

If this were used for blood glucose, the blood would have to be extracted first by traditional means.

This does not seem to be the intended use, because the already existing invasive devices are more convenient.

The intended use appears to be for testing food using a device that contains the gel and to which you attach the smartphone in order to read the results.

Yes, I understand how they conducted the experiment! Because this would not be place under the skin, it still constitutes a noninvasive sensing technique
Distinction without a difference, your lead comment is a list of common issues for something that is orthogonal to the article. I appreciate it regardless but some humility is in order :) not being strictly on topic is fine, it's still interesting. But I don't think arguing they're the same thing repeatedly down thread is kind.
If this were used for blood, it would be invasive, because the blood must be extracted with needles, exactly like in the current kits for measuring blood glucose.

The gel holder would substitute the sensitive strips that are used in such glucose measurement kits and the smartphone would replace the electronic reader that displays the results.

Nothing would change about the invasiveness of having to pierce the skin with needles, to extract blood.

There exist experimental devices for non-invasive measurement of the blood glucose, whose purpose is to avoid the use of piercing needles, but this new glucose sensor has nothing to do with those.

This new glucose sensor is for other uses and its advantage is that it no longer needs a special expensive electronic device for reading the sensor output, but a standard smartphone is good enough.

(comment deleted)
> 3. You're taking medication that "looks" similar to glucose (in electrochemical sensors that's ibuprofen, but the number of magnetic analogues, especially in the presence of interference, is surely greater)

Huh, I never knew this. Does this mean taking ibuprofen can give you an elevated glucose reading on a blood test?

I misspoke, it's actually acetaminophen (I'm always confusing the two!): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5950984/

But yes, presence of acetaminophen with a naive electrochemical sensir will cause false high readings, which are VERY dangerous!

The latest generation of sensors have fixed this problem, but with any new sensing tech, you need to make sure you're learning from the mistakes of the past. With a new sensing paradigm, you won't necessarily have problems with the same drugs, but knowing that this kind of constructive interference is a potential problem will help you design a more robust system.

Just a note about false readings. A diabetic with high glucose is deeply uncomfortable, and will suffer long term damage, but is not in immediate peril.

The bigger danger is being low, which will lead to unconsciousness, coma, and death.

Insulin brings down blood glucose. Getting insulin when not needed can be deadly. (It's a favorite of murder mystery writers.) False highs are more of a risk than false lows.

Which means you may need to administer sugar to a diabetic in an emergency. But unless you're a medical professional you will never need to give them insulin.

Absolutely, the danger of a false high is that you dose insulin to correct and end up pushing yourself into a dangerous hypoglycemic state. That's why acetaminophen interference is so important to deal with! And while it's true that running a bit high (250-350mg/dL) isn't an immediate risk, DKA is still a possibility once you breach the 300s, and that will land you in the hospital.
I’m more interested in measuring insulin than glucose.
>Similarly, the team’s study suggests that a cellphone magnetometer can measure pH levels with the same sensitivity as a thousand-dollar benchtop meter but at a fraction of the cost. A home-brewer or a baker could use the magnetometer to quickly test the pH of various liquids to perfect their craft, and an environmental scientist could measure the pH of ground water samples on-site with higher accuracy than a litmus test strip could provide.

To determine if my body is burning ketones on a ketogenic diet, I used to have to pee on pH-sentitive keto test strips. Sounds like in the future I'll be peeing on my phone instead.

Doesn’t a electronic PH meter work for this purpose (hopefully not shared with the kitchen)?
This doesn’t look like the breakthrough the headline suggests, at least for glucose monitoring.

They created a material that contracts/expands depending on measurement levels, and that can be detected by the magnetometer. But the sample needs to be deposited in it.

This is not a non-invasive glucose monitoring solution, and current under-skin devices will remain a lot more convenient.

A magnetometer costs $1-$5, plus a couple dollars for a microcontroller, what will likely happen is that this will result in a new breed of devices that offer laboratory precision at very low cost. The phone part is irrelevant.

Of all the things they described as possible, measuring glucose is the least interesting, since we already have a better tech for that.

They have a number of other potential applications, and it's rather clever and surprising. They could have led with any of those.