How do you get a decent ECG with a wearable that has a single point of contact with the skin. If you need to touch a contact on the device to get a two lead equivalent ECG then how can this be real time monitoring?
Continuous monitoring is extremely challenging still because ECG data needs to be sampled at a relatively high frequency (~200 Hz) to accurately identify the QRS complex in the waveform. That uses a lot of power, and the batteries we have still aren't good enough to support those types of demands. 200 * 60 * 60 = 720,000 samples per hour to collect and process.
It's possible that algorithmic approaches may be able to reduce the sampling frequency required. Power-constraints were the main issue when I studied this topic 10 years ago during my master's degree. I had looked into non-frequency domain techniques (such as empirical mode analysis/Hilbert-Huang transform) as a possible way to reduce sampling frequency and thus power consumption.
https://github.com/AshwinSundar/Empirical-Mode-Decomposition...
rather than reducing the frequency of the sampling, dynamically adjust the duty cycle of when sampling is happening?
this is probably a dumb suggestion, it seems pretty obvious. for example the apple watch doesn’t do o2 monitoring continuously, just for some fraction of the time.
do you need to sample every second to detect heart attacks? don’t they continue to show up on an ekg for more than 30 seconds?
Making it effectively useless? Unless the fraction of the time is multiple times per minute? E.g. in sleep apnea it's not uncommon for some desaturation to occur, triggering an arousal and deeper breaths, restoring saturation, only for the cycle to repeat 2 minutes later.
My Garmin has a similarly useless feature. I have no idea what the supposed benefit is. Maybe they hope that if they sample multiple nights they can detect some desaturation anyway and can get the user in for polysomnography? Might be worth it.
With my cpap, I normally have two 1-2 events/hour but my blood oxygen no longer drops
I went through this with a relative just two weeks ago and learned this from the cardiologists
Integrated frontends have had big impacts on efficiency and improving batteries have had basically the same increase in specific energy. I would be shocked if power was a limiting factor.
https://www.ti.com/lit/ds/symlink/ads1291.pdf?ts=17469713549...
Detection of ecg anomalies(especially episodic ones with intermittent recording) was the subject of the physionet cardiac computing challenge almost 10years ago[0].
It’s amazing how far machine learning has come. I know teach a version of this challenge as a one day in class activity in my department’s physiology class. They actually get to train multiple models on a gpu cluster (and compare that to trying to train models on their laptops).
One thing we reinforce in the lesson is human vs. computer “interpretation”. They/clinicians can look at ecgs and make some sense of them. An LTSM is worse than random chance/a medical student. However moving to the frequency domain makes the LTSM more accurate than cardiologists, but neither they nor clinicians can “see” afib ina spectrograph. It’s a great way to talk about algorithmic versus human reasoning and illustrate that to students.
That then gets reinforced with other case studies of the ying and yang of human and machine decision making throughout our curriculum- like alpha fold working great until you ask it about a structure in the absence of oxygen, because that’s not in its training data.
But to be clear, a single lead ECG requires two electrodes at a minimum and commonly a third as ground. So a single lead ECG will have minimum two cables attached to electrodes on the patient. The placement depends on which lead (eg lead I, lead II, etc) but there's always two minimum.