Electronics Teardown: Stelo Continuous Glucose Monitor (2024)
andykong.org
andykong.org
> While GOx maintains this level of activity in vitro, its stability in vivo is of the order of 10–14 days, necessitating novel immobilization and enzyme modification strategies to extend the functional lifetime of these oxygen-sensitive sensors.
https://pmc.ncbi.nlm.nih.gov/articles/PMC3879770/
Coincidentally, when I was doing this field of research, I wanted a hardware/software design that could algorithmically compensate for the sensor degradation. I even wanted to mathematically model the electrochemistry to find a clean solution. My electrode was nanoporous platinum, which then was considered more advanced/fashionable/versatile than glucose oxidase coating. In the end, I decided on reading off some cached table, not unlike a consumer electronics estimating remaining battery life.
On a separate note, you can design a fuel cell out of blood glucose. So, your body glucose generates electricity to fuel your electronics. The amount of electricity generated is not too off the target power for ultra-efficient circuits, and we even conducted in-vitro experiments as a proof of concept.
All of this was 20 years ago. I am glad that there are commercial solutions now but a bit saddened that the field hasn't progressed as fast as other areas of technology.
My (weakly held) hypothesis is that ML stack on IR already maxed out, and there is no more signal we can get out of. H20 interference is still difficult to work with. Raman is the way to go for those cases. An affordable, portable (true) Raman could change how we do medical and chemical sensing in aqueous solutions, anywhere from medicine to food production.
I actually spent a full year outside academia doing IR v Raman work. I ended up deciding on industrial IR with a sophisticated signal processing & ML stack, but at some point, I was planning to build a Raman from scratch because I was so tired of compensating for O-H bonds.
I also made another comment under this post about IR v Raman work I did. That work I have done is still relevant in the field, but I am hoping someone comes up with a better solution.
Thanks for your kind words!
Type 1 diabetic here: one of the reasons for changing the insertion site is because of the risk of infection / rejection. It's not just about the power requirements.
As an alternative, Eversense is a long term CGM inserted by a healthcare provider that's supposed to last a year. I've heard mixed reviews.
Speaking of mixed reviews, I have mixed feelings about non diabetics using CGMs. But if it lowers the price for all of us, then I'm on board.
What downsides do you see?
e.g., I feel sad when a non diabetic uses a CGM and says something like "omg I had a milkshake and my blood sugar spiked all the way to 120!". There's nothing wrong with them saying that, of course. It's just challenging to hear when you have a body that can't produce insulin at all.
- If more people buy something, in the short term, you have shortages.
- In the long term, you have better economies of scale and prices go down.
These things should be available for maybe $15 at most, generically, off-brand.
Personally, I think this is the kind of exercise most people should go through perhaps annually.
Thanks for the awesome video.
It's not the wire that's precise here - it just provides enough force to get a the stamped metal piece over the tooth of the plastic gear, then it slams into a hard stop, then the button batteries dump current into the other wire, which pulls it the other way, pushing the gear one tooth width, and slamming into the other hard stop. Each back and forth ratchet of the tooth turns the worm gear on the plunger the precise 0.05 units.
Another cool part of these is the initial insertion - there's a huge torsion spring that stabs a hollow needle into the skin then fully retracts it, leaving the rubber tube.
Then there's a little corner of the PCB with a cutout and power trace, you can stab with a paper clip to disable the pod in an emergency.
Then there's this membrane that equalizes the air pressure in the pod if you're flying with it. It lets air through but not water. So you can dive in a lake with it too.
I love clipping these things open and just marveling at the clever engineering for cheap mass production of such a sophistocated device.
Regarding price, my point still stands that wire is not cheap enough to motivate its use. Your link actually proves my point (1.6 USD/30cm). Pager/toy motors are costing cents.
But maybe it is more expensive. Then they probably choose it for the reliability, and a simple and quiet design. Or perhaps the button batteries don't have the energy to drive a pager motor.
While the pump manufacturer just puts a 1000m spool in his assembly line every few hours.
If you can design the wire to require 3x less material than the motor approach, it will win.
This likely why the sensor life is advertised as ~12-15 days:
"*Sensor Survival Clarifying Statement: A study was conducted to assess the sensor life where 77.9% of sensors lasted the full 15 days. In other words, when using the product per the package labeling, approximately 20% of sensors may not last for the full 15 days, 10% of these sensors may last less than 12 days."
https://www.stelo.com/faqs/using-stelo/how-long-can-i-wear-m...
https://idoroseman.com/freestyle-libre-blood-glucose-monitor...
> The sensor life is limited by the amount of chemical coating on the probe and the amount of glucose in the interstitial fluid. It could last longer for a person with low and steady BG than for one with chronically high, so the sensor life is a conservative compromise between battery size, Bluetooth wake cycles, and probe life. If the probe had a larger supply of reactant or its dispersal rate could be slowed, the sensor life could be extended to the battery capacity or the max that the FDA would permit.
https://sequenex.com/the-otc-cgm-market-comparing-stelo-ling...
> Stelo, created by Dexcom, was the first of these sensors to receive OTC approval from the FDA. This small, completely disposable device uses the same hardware as the Dexcom G7. The main difference between it and the prescription-only device is the software that pairs with it. While the sensor has the same capabilities as the one marketed for people on insulin therapy, the OTC software restricts much of this functionality.
> Lingo.. uses the same hardware and housing as the [prescription-only] Freestyle Libre 3. Like Stelo, the software system is where the real differences exist... Abbott’s second OTC CGM is Libre Rio. This sensor looks identical to the Lingo and Freestyle Libre 3. The main difference is in the marketing
That is really unfortunate, because the default Dexcom apps are terrible (the Stelo app is slightly better / more modern, but has all the limitations of the Dexcom app, plus no calibration).
xDrip's got 'em figured out and they'll happily fire up without an issue there.
The only down side is the software states it's not for t2d on insulin or 5qd for that matter. It only shows input from 70-240, outside that range all you know is it's above or below. So it's hard to tell how much correction you need at times.
It really sucks how much medications and cgm cost over the counter.
Wouldn't 1 per 5 minutes suffice? Just a couple of bytes should be fine for the reading. What other data could possibly be needed?
To me this smells of it sending a huge json payload with software version numbers and other unnecessary junk simply because the interface between the hardware and app wasn't designed with power efficiency in mind.
This post about the Stelo is fascinating.