Building an Open Source Blood Pressure & Heart Signal Monitor -- Episode 674

The above-the-elbow ones are considered more reliable and are more common, but funnily enough, signal-wise, the wrist ones have a clearer signal compared to the above-the-elbow ones, because there is less muscle and fatty tissue near the wrist compared to the biceps area!

Really interesting, I'll look into how that works!

"...where an additional cuff would be placed near the wrist to monitor sounds..."

Just to add... I have both an arm cuff monitor and a wrist cuff monitor. Even though I had seen reports suggesting that the wrist cuff is a less accurate method, I have found it to give pretty consistent readings when compared with the arm cuff one. You have to be mindful to raise your wrist up to chest height whilst taking the reading though.

"...Haven't seen ones with the clothing detection..."

The earlier version of the 'Omron M6 Comfort' monitor has it

In my eyes, the only kind of trend that can be established is, as you've said, by routine monitoring at home. Having those one-off measurements sounds rather unreliable and would only serve the purpose of checking the current condition of the person if they are feeling unwell at that moment. As you've pointed out, too many factors influence the pressure reading.

Haven't seen ones with the clothing detection, and honestly wanted to try over thin fabric shirts since to me it doesn't look like it should cause any errors in the readings (if it's light material and it's not constricting blood flow through the arm). I'm impressed that they can detect pieces of clothing. I'll need to explore further how they do that. Sounds cool!

The stethoscope I had here was broken, so the readings were bad, but that's something I wanted to try, to see whether the manual method could be recreated like that. There are some other fun methods I've come across while researching, one of them being a two-cuff method, where an additional cuff would be placed near the wrist to monitor sounds, and with PPG, which is what I tried here. 

A lot of cool research and devices have already come out to estimate blood pressure without a cuff. Some principles include looking at the delay between the ECG and PPG signal, which essentially shows the speed of the mechanical wave through the vessels, and other work solely by looking at the morphology of the PPG signal. So there are a lot of fun things to explore and research about this subject!

"...so I guess the general idea is to establish whether there is a clear pattern of constant high or low pressure..."

This is why I'm generally sceptical about one-off readings taken at a doctor's check-ups, often taken years apart. 

"...and that's the white coat syndrome..."

Another reason why I'm sceptical. You could have low blood pressure but if you also have white coat syndrome, the reading could appear within the normal range at a check-up.

Add to that, that check-ups usually occur at different times of the day and don't factor in activity, food, stress, etc. on each occasion.

Realistically, I think a clear pattern can only really be achieved by routine monitoring at home. The individual needs to collect sufficient data to be able to present to the doctor.

Interesting findings on the cuff inflation. Another mystery is how some of them do clothing detection. My monitor will throw a clothing error if the sleeve of the t-shirt gets caught under the top of the arm cuff. 

Yes, getting different results from the same measurement by using different percentages raises some questions. Perhaps need a cuff with stethoscope built-in so as to get closer to that of the gold standard manual method.

Thanks, glad you enjoyed it!

Even one minute apart, the measurements will be off, especially since all of the non-invasive procedures are just estimating the pressure. Another thing is, pressure by itself is something that, of course, drastically changes throughout the day, based on activity, food, to stress, so I guess the general idea is to establish whether there is a clear pattern of constant high or low pressure. But there is another issue here, and that's the white coat syndrome, which can actually go both ways! Some people get extra nervous when they are at the doctor and their blood pressure jumps up because of that, while other people feel at ease when they are at a doctor and their blood pressure drops.

Since this was all inspired by my grandma, we always make sure she sits for at least 5 minutes, and if it's not a crazy high value, we remeasure after 10-15 before she takes any additional medication to lower the blood pressure, and usually, the pressure is much better the second time around!

I haven't done any experiments in regards to estimating the systolic pressure based on the cuff inflation, got curious so I extracted that data from the video dataset. This is what I get:

The data is much noisier because of the pump, which is to be expected, of course, but pulsations can still be seen even after the systolic pressure has been surpassed. One reason why pulsations can still be seen is that the cuff, while cutting off circulation there is still in contact with the upper part of the arm and transferring that mechanical wave. I am not sure how they reliably cut off at about 30 mmHg above the systolic pressure, but as you've mentioned, it's rather reliable. Sometimes it fails and needs to be reinflated, but rarely.

As for how long the deflation lasts, there are probably some models that can vary that speed as well. I remember finding a proposed mmHg/s figure in some research paper, to which I adjusted the valve I had. The ones I've taken apart have all had a small adjustable valve. Having a fully closed system with two valves would be an even better option, since that would allow for more complex and accurate algorithms where the device could keep the cuff at a set pressure level while conducting a measurement over a longer time. The time difference could also be attributed to how tightly the cuff was put around the arm, since that affects how much air will actually go in the cuff, though I don't know how much time-wise that would affect the deflation phase.

My mistake for mixing up the 40% and 50% sorry, and here is why ahhahaha! Two of the more popular methods are the Sapinsky and Geddes methods, and they both work in the same way, look at the envelope of the signal, and using the percentages, they do the pressure estimation. One of the methods proposes using 40% and 70% while the other proposes using 50% and 80%. This is something that I found weird when looking at it since this meant that we can get two rather different results from the same measurement. I had the methods mixed up in my head a bit while doing the video, so I've probably used one in the beginning and one in the end!

Another impressive project.

I have to wonder though when was the last time a doctor/nurse made someone sit still for 5-10mins before taking a blood pressure reading. I rarely get the same reading between two measurements taken sitting quietly 1 minute apart at home, so am sceptical about doctor's check-up measurements often taken months or years apart.

How is the rough estimation of systolic pressure done during cuff inflation ? I have found it pretty reliable and as you point out usually around 30mmHg above the final systolic reading. Very occasionally does the pump turn back on to inflate the cuff more. Is it just monitoring for loss of heart rate detection and then adding an extra 30mmHg ? It appears to lock onto heart rate through the cuff pressure pretty quickly at about 5 seconds/40mmHg into the inflation. 

Interesting insight into the two valves. Sometimes when taking measurements I've noticed that the measurement deflation phase appears to take a lot longer than others, even without significant variations in final readings. Up until now I had assumed that it was just the EM valve controlling the deflation phase but it looks like the adjustable valve and signal monitoring is doing the work, until the final cuff pressure release.

In the overview section you mention 40% of the envelope max for the systolic pressure time

however later on in the analysis you start to use 50%. Was this adjusted after experimentation to more closely match the reference readings from the digital gauge ?

Thank you very much, glad you liked it!!!

Wow.  Impressive and thorough.  Excellent job!