Living with Type 1 diabetes is a numbers game. There’s not a moment in the day free from the burden of tracking your blood glucose concentration, making “What’s your number?” a constant question. Technology can make that question easier to ask and answer, but for T1D patients, especially the kids who the disease so often impacts, all that tech can be a distraction.
To solve that problem for his son, [Andrew Childs] built this custom T1D smartwatch. An Apple Watch, which integrates easily into the Dexcom CGM ecosystem, seems an obvious solution, but as [Andrew] points out, strapping something like that on a nine-year-old boy’s wrist is a recipe for disaster. After toying with some prototypes and working out the considerable difficulties of getting a stable BLE connection — the device needs to connect to his son’s iPhone to get CGM data — [Andrew] started work on the physical design.
The watch uses an ESP32-S3 on a custom PCB, as well as a 1.69″ TFT IPS display and a LiPo battery. The board also has an accelerometer for activity monitoring and a vibrator for haptic feedback. Getting all that into a case was no mean feat, especially since some degree of water resistance and shockproofing would be needed for the watch to survive. [Andrew] had a case made by a local 3D printing company, and he managed to source custom-cut and silkscreened glass for the face. The result is remarkably professional-looking, especially for a software developer who hadn’t really stretched his maker wings much before tackling this project.
[Andrew] doesn’t appear to have made build files available yet, although he does say he intends to open-source the project at some point. We look forward to that as it’ll be a big help to anyone trying to hack diabetes care. Until then, if you need a primer on continuous glucose monitoring, we’re happy to oblige.
If the final goal is a commercial productt and not a “because I can” project, why not cooperate with a company that has experience in this area? E.g. Lilygo sells smartwatches with an espresssif chipset. I haven’t checked the details, but their hardware seems a good fit for this project.
Selling medical devices will be regulated by the FDA/Health Canada/Your country’s equivalent. Then there are the liability issues. This has to be a DIY sort of thing. (Google #wearenotwaiting for more DIY diabetes things.)
Acually Andrew states that he’d like to open source the project. The Lilygo t-watch might be an excellent fit for this as it should require relativley little change to the code and would open it up to lots of people who might find the hardware side a bit intimidating.
Interesting read. I don’t get where Https comes into it though. If the watch is just using ble, it could use a much more lightweight processor such as the Nordic nrf ones.
Probably all that apple will allow apps to access and Andrew needed some way to keep the Continuous glucose monitoring (CGM) data private.
Dexcom (the provider of the CGM) allow third parties to connect directly to their servers to get your data. I assume this basically uses the phone to access that, rather than getting data off the CGM directly.
I use the Dexcom app on my Google Pixel 2 and it’s ….ok. I’d really like to have a different interface though, and this of project might inspire me to have a go if the source is opened up.
and sending all data to china!
If the Chinese can use the data to improve the lives of Type 1 Diabetics I’m all for it.
how?
I wonder how the battery life is. Lilygo’s TWatch series gets maybe 8 hours with BLE going, and it underclocked.
ESP32S3 is pretty power hungry for a couple hundred mah battery
He dedicated a full paragraph to that in the article. Said that most of the watches he assembled got 3 days but his son’s gets 6-7 days
Im guessing this isnt invasive sensing needing a poke into the skin.
I had the same thought. Two questions arise: since both methods require a smart phone the plastic disk implant reduces parts and what do we know about the relative accuracies of the two approaches?
Well, mainly exact numbers aren’t so important as trends with large swings, couple points different between sense methods is not important, or at least for me and my version. The injection takes a while to come online anyways so shooting up is a kinda guess
Still no answers. How does sweat affect their accuracy? I never said “small swings” I am in fact wondering about large differences in accuracy between the two devices. If one is going to be very wrong most of the time that’s not the one I want. And how would it work without pricking the skin? Is this similar to the IR pulse ox finger clip?
The watch is displaying data that is sent to a phone from a Bluetooth glucose sensor. These sensors have a small sensing filament that gets poked into the skin when you install it. They last between 10 an 14 days.
Sounds painful. Explains how some of the glucose watches work though.
Very nice project, I really like how much attention to detail he put into the final version, even optically bonding the custom cover glass. Fun fact: There finally are quite a few smartwatch-style AMOLED displays on the market that can be driven via SPI/QSPI.
I was trying to figure out why an Apple Watch wasn’t a good fit for, then I read Andrew Childs’ blog (issues include distraction at school and making sure the bluetooth connection is persistent). Very cool project. Also, not all CGM makers have a watch app.
Yeah, it’s not the readout device that’s the problem here, it’s the sensor jammed in to the body. As soon as I saw it I walked away.