Detecting heart arrhythmias using machine learning and Apple Watch data
insighthealthdata.comYancheng was awesome to work with. We + UCSF actually just launched a continuation of his work via the mRhythm Study:
https://mRhythmStudy.org
I installed the app and I got a spike of 121 bpm between 4am-4;30am for the past couple of nights. That seems a bit strange. Nightmares? Buggy software?
Getting some spikes is normal while you're in REM sleep. Here's an example of restful sleep vs sleep after a night of drinking: http://imgur.com/l9RVeK5
In general, we'd like to add more comparisons in the app so that you know how your stats fit in to the larger population. Stay tuned. :)
I predict that in 10 years, more people than not will be wearing a smart watch of some type that will continuously be tracking vitals. You'll probably even be offered a free one by your employer or health provider.
Average heartrate gradually going up over the last few weeks? Your doctor's office will ping you to drop by. Sleep patterns have been erratic lately? Send down an electronic questionnaire to see what's up. You see? If you unveil the symptoms sooner, you can treat them sooner. That makes sense. $$$.
That's a dystopian scenario, given that poor health is often an excuse for discriminating against the unfortunate people who experience it.
Disclosing everything to a trusted employer or health provider is a step we need to be very careful about. It's panopticon territory.
Also I bet we are going to find the health statistics of people found to have minor arrhythmia through cheap continuous screening are going to be completely different than those that come in with a complaint and are given a one-time screening. It may turn out to not be as clinically useful as hoped.
Almost certainly true. Also, how accurate are these optical HR sensors?
Chest strap monitors are somewhat more accurate than optical monitors:
http://www.wareable.com/sport/heart-rate-monitors-chest-stra...
And it's unclear whether the optical measures are accurate enough to inform health decisions. I can say we've looked at this at our company and it's a fairly complicated picture...
I doubt it. Surely I'm not the only person who finds watches annoying to wear. I'll only start using these devices if I'm not constantly aware of the thing.
Aww too bad I have a Microsoft Band2 not an Apple watch.
I'm watching my heart rate like crazy since I was diagnosed with hypothyroidism last Friday and my metamorphosis back to being human again since I started taking synthroid it's interesting to see but especially my heart rate.
My resting heart rate which was around 70 bpm (yeah I'm a slug) but now it's about 85 bpm.
The Microsoft Health website to view Band2 data is OK but not really detailed e.g. my heart rate last night https://i.imgur.com/E4Y8IzZ.png
I only wish I had the Band2 or any fitness tracker sooner to get more of a baseline or better yet before my thyroid decided to go on it's five year death spiral and take me with it.
I'm hoping with these gadgets at some point we'll have enough data tracked to do meaningful analytics, if not just on yourself but across a lot of voluntarily uploaded data. Not a big deal with a heartbeat every 10 minutes right now, But 5-10 years from now that would at minimum be very cool.
My feeling is that for health and safety reasons everyone will soon be wearing smart watches:
https://h4labs.wordpress.com/2015/07/28/in-the-future-everyo...
We could use a few more sensors like blood pressure and skin temperature.
What's really needed is for Android Watches to aggressively compete with Apple. Apple and Samsung, for example, would really push each other.
The MS Band2 has skin temperature and galvanic sensors but you don't see the temperature data like heart rate. I think skin temp used for the calories burned data but it would be nice to see it, there are third party apps to view it. It has a UV sensor too but only on-demand when you select it.
No I don't work for Microsoft.
Blood pressure would be tricky then again the watch is around your wrist so maybe not.
A real-time blood sugar reading would be nice to see too.
According to a supplier of sensors, OEMs are currently lining up electrical resistance sensors for their next smartphone models. Skin temperature is still too hard for smartphones, on account of the device's temp fluctuations.
Look out for fun lie-detector apps coming to your app-store-of-choice.
and peeing into toilets with live pee-anaylsis monitors.
Eventually this will be a great tool, but today, I would not put my well being into a smartwatch, with or without machine learning