Atrial Fibrillation (AF) is the most common serious heart arrthymia, affecting around 30 million people worldwide. People with AF have a 4 times higher risk of mortality and 5 times higher risk of stroke than the normal population.
I've implementated a widely-cited CNN architecture for ECG-based AF classifcation. Classification can be run directly off a 60 second recording from a Polar H10 Heart rate monitor.
Hi stenmorten try going into Settings > Privacy & Security > Bluetooth, and adding your terminal application or development environment to the list of applications that are allowed to use Bluetooth
The blue circle is your chest expansions and so when it follows the gold circle you’re controlling your breathing well. Good hrv is above 150 ms and is shown as a green band on the bottom graph
Yes basically, I remove the gravity component by high pass filtering, and then take just the component of acceleration which is along the axis of chest expansion (i.e. pointing out from the sensor) off the top of my head this is the z axis.
But this only works when you are sitting quite still, movement introduces a lot of noise to the acceleration
It currently only works with a Polar H10, which has a heart rate monitoring (ECG) and an accelerometer (for measuring breathing rate through chest expansion)
That article is referring to the internal memory that the Polar H10 has, which can be used to store some data. But you will eventually need to connect the device to retrieve that data. You can access this data programatically with a blue-tooth-low-energy client like my application is doing (using python)
This would be possible I with the Polar H10 I think. The H10 sends raw ECG and analysis on this waveform could be used to find anomalies like afib. I haven't done this myself though.
This is measured over a very short time scale though – so not a persons absolute max and min heart rate, just how it varies over a few seconds. And yes, higher HRV is generally healthier and points to a better balance between the fight-and-flight response, and rest-and-digest response.
Yep that's right, and the H10 is able to stream the ECG waveform. For this project I used the inter-beat-interval that is sent however, as it means I don't have to run peak detection on the ECG to get heart rate.
The H10 is a newer and better version of the H7. Because breathing rate is measured from chest expansion that is pretty easy to verify (there can be noise in the estimate of course due to movement artefacts). Heart rate estimation from the H10 is widely considered to be reliable, The H10 also offers to ability to stream the raw ecg to check this.
Yes it's common to estimate breathing rate based on the oscillations in the heart rate pattern. But estimating it from the chest expansion is a more direct method
Interesting, thanks for sharing. It's cool they do it as a platform, but I'm not sure what the benefit over using a Polar H10 would be, which are widely available, I guess it would be custom branding of the hardware
I've implementated a widely-cited CNN architecture for ECG-based AF classifcation. Classification can be run directly off a 60 second recording from a Polar H10 Heart rate monitor.