▲ 142 points
back
29 comments
OP here.
Many of you have helped out with the mRhythm Study we launched a couple of months ago:
https://itunes.apple.com/us/app/cardiogram/id1000017994?ls=1...
We're planning to start publishing some of the insights we're learning from the study, and we thought we'd kick off with a post on the foundations:
* How normal rhythm varies and why an "irregular pulse" is actually normal
* What optical heart rate sensors measure, and how they're different from the ECGs commonly used in hospitals.
* An example of an "irregularly irregular" abnormal heart rhythm.
* An example of a regular abnormal heart rhythm (atrial flutter)
* An answer to one of the most common questions we hear, "Could Apple Watch detect heart attacks?" (Short answer: not yet.)
If you have questions or things you're curious about for the next few posts... we'd love to hear your thoughts!
I know this is probably not the feedback you are looking for, but: your page at http://www.mrhythmstudy.org/ uses a non-standard font (Avenir Next) which doesn't show up in Windows; instead what I see is an ugly Times New Roman everywhere that makes the page look like it comes straight from the Geocities era. Use something like Google Fonts to embed a custom font to fix it so that every client downloads a copy of the font before displaying it.
Thank you! That is helpful feedback.
As a med student/programmer, I thoroughly appreciate this write up. Excellent job, OP.
A funny aside: many of my classmates own Apple watches/Fitbit HRs. I've heard of at least one student looking back at her heart rate data from finals week and saw spike to 120 bps that was maintained throughout one of our more stressful 8-hour exams. Her fitness tracker registered it as "exercise."
Having been in A-fib several times in my life, if you ever feel like your heart rate isn't steady and isn't changing in the usual way as a result of exertion, GO TO A DOCTOR IMMEDIATELY. Lesson learned. If anyone ever releases a watch that can identify conditions like this with a great deal of specificity and sensitivity, I would consider that far more of a killer app than email notifications on my wrist :).
You might even call it an anti-killer app...
Health monitoring is going to be incredibly helpful. Beyond just pulse monitoring, I'm looking forward to analyzing trace materials in sweat and the like to notice deficiencies or issues.
Just to be clear - A-Fib needs medical attention, but it probably won't kill you directly.
It's possible to have permanent A-Fib and survive - if not thrive - just fine.
The three things that can kill you indirectly are clots created by blood pooling in the heart, heart exhaustion from tachycardia, and a heart attack from possible ventricular arrhythmia.
I occasionally get A-Fib and it's consistently caused by stress - often lack of sleep (no more all-nighters...) or any situation where I have too much to do and not enough time to do it.
(That thing where people say "Hard work never killed anyone?" 100% wrong.)
Having researched it, the causes are complicated, and stress is only one predictor. There are physiological and genetic predispositions and other possible contributing factors.
So yes - GO TO A DOCTOR. You probably won't die - but you might, and it's very easily avoided.
That looks like a very interesting write-up about ECG, and I'll read it properly tonight when out of the office - thanks for writing!
I'm a Garmin Edge user of many years, always interested in HRM during activities. But that's not something I wear all the time. Hopefully when my Pebble Time 2 arrives something similar can be done on that platform.
I'm intrigued to see what sort of details may be captured from such a seemingly simple device.
I've got all kinds of gaps in my cardiogram which could apparently mean atrial fibrillation, but I suspect it more means that my watch is just crappy at measuring my heartbeat. Sometimes when I'm out on a walk it says 60-70 when I know I'm at 130-140. Not sure how to make my watch more accurate, it's already strapped on pretty tight.
That's part of what we're trying to quantify with the mRhythm Study--both the accuracy of heart rate sensors on average, but more importantly, how to predict in which situations and for which people the sensor is likely to work accurately enough for clinical use.
One tip I've found: if you wear the watch slightly higher up your forearm, it often gets a better reading than if you wear it directly on your wrist.
I have the same issue, seems to be very erratic with its measurements. Frustrating as fitness tracking was one of the main reasons for getting one in the first place.
I would also like to be able to up the number of measurements it takes. I often only use about 50% battery a day so would happily utilise the remaining charge for more regular monitoring.
I sullied my apple watch by wearing it while playing hockey for science. AMA.
I lost 3 fitbits that way! That clasp was mostly useless for anything more active than running.
Does the HR graph reveal stages of grief, I mean spikes, as you realised the damage?
We have looked at my wife's data when she had a panic attack a while back. You could almost see the exact point it started as well as when it tapered off.
It was interesting in her case to see that her heart rate jumped well before she thought it did, and that the actual "trigger" to her attack may not have been what she thought it was.
That's very interesting about physical triggers happening before the perceived start of the incident.
Agoraphobia? Best wishes for working through that, it certainly takes a lot of time but you can get there (or at least to a state where it doesn't happen as often).
Looking forward monitoring via Apple Watch not just user’s iPhone battery with the app like this: https://itunes.apple.com/us/app/watchmybattery/id1024068907 , but the gas-tank
i thought you needed three electrical connections for ECG?
This is not an ECG. It is simply looking at your pulse over time.
Any circuit through the heart is enough for (at least a single-lead) ECG. For example, the AliveCor is a hand-held mobile phone case—when your left hand and right hand are both touching it, you can get an ECG.
Chest ECGs done at the hospital are often 12-lead, and those give different "angles" through the heart, which gives cardiologists more clues on exactly which parts of cardiac tissue are causing a problem.
Well the 12 lead is more finnessed than that - essentially without a 12 lead your ability to definitively rule out an acute heart attack is completely impaired. 5 lead ECGs will still miss (some high percentage, above 10% and below 50%) of heart attacks.
A one (two?) lead ecg as you describe is not going to tell anyone what is going on with ST segments, particularly as you would be only looking at your aVR or aVL(depending on which way you sum the currents).
Yeah, but anyone who's trying to dx (or, worse, rule out) an AMI on the basis of what their Apple Watch can report is ... yeah, I don't even know what they are. Not the sharpest pencil in the box, for a start.
The augmented leads (aVR, aVL, and aVF) require a minimum of three electrodes (multiple electrodes are connected to ground).
An electrode on each arm gives you Lead I.
Lead I is not as useful as Lead II, but it's still plenty useful for diagnosing all sorts of cardiac issues (there's a lot more that can go wrong with your heart than just an MI).
Presumably a single lead would still be able to detect certain arrhythmias? I.e. if there are no QRS complexes, then there may be something wrong with the person wearing the device. There are benign explanations for why the signal might be weak but also lethal causes like Vfib -- with a few additional, I don't think it's implausible for a device to be able to make the decision to ask the user "hey, are you okay?" and then potentially seek help if the user doesn't confirm that they are OK within some set period.
Absolutely. Consider the fact that an AED only has one lead to work with, and it is able to recognize a couple different lethal arrhythmias in order to deliver a shock (and more importantly, _not_ deliver a shock at any other time).
To be clear, by "lead" here, I mean a pair of electrodes. A single electrode (like you might put on a watch) is useless.
12-lead ECGs are really only done with 10 physical leads. One lead is ground, six leads are each single-channel measures (V1 -V6) that wrap around the heart and the remaining three leads are used to create six channels - three single-channel measures, and the last three channels are calculated as diffs between them. V1-V6 have specific spots on the chest to live, but the remaining four leads (including ground) get one limb each, and generally work wherever you want on that limb.
(I used to be an EEG tech, and for us, the ECG was the primary thing we wanted to get rid of...)
I'm actually really curious -- why would you want to get rid of the ECG as an EEG tech? Do the signals interfere? (I don't know too much about EEG -- happy to learn more!)
The short form is that EEG is brain and ECG is heart, and you only want to see brain waves, not heart waves - it's pure interference, that changes the shape of the wave. It's simply 'not the thing we're looking for'.
Most clinical EEGs are related to epilepsy or similar, which usually has a particular pattern with spikes in it, but sometimes that pattern is subtle in the EEG. The spiky nature of ECG 'bleeding through' can make it a bit harder to determine the true EEG spikes from the artifact ECG patterns. Usually you will record a single-lead ECG along with the EEG, so you can see where the ECG spikes occurred and account for them in diagnosing the EEG.
This link has some clear spike-and-wave epileptic waveforms in it, but it's not always so clear-cut: https://en.wikipedia.org/wiki/Spike-and-wave
Finally, there's not a lot of overlap between neurological and cardiac patients - there's rarely any need for an EEG tech to do a full ECG (we never did, and we did approx 3200 patients/year), and full ECGs are pretty commonly done in most other places in a hospital anyway.
Apple does not save lives. Neither does their watch.