Sleep 3.5 Factors Affecting PPG Signals

Learn various factors influencing the accuracy of PPG signals, including individual variability, physiological factors, and hardware differences.

Video Breakdown:

00:11 Factors Affecting PPG Signals

00:47 Study on PPG-Derived Heart Rate and HRV

01:35 Impact of Motion on PPG Signals

02:19 Device Performance Comparison

03:21 Skin Tone and Heart Rate Accuracy

04:24 Oxygen Saturation and Skin Tone

08:17 Practical Recommendations for PPG Use

Now, the final part of the discussion I wanted to talk about is sort of the factors contributing to the PPG signals.
And these include, various factors here, and there are many sources of variations of PPG. This includes individual variability, including skin tone, obesity, age, and gender. There are physiological factors, some of which I've mentioned before, such as respiration and venous pulsations, body site, and body temperature. There's external perturbations such as motion artifact, ambient light, applied skin pressure, and also design, differences in hardware and software inconsistencies that occur, across the different devices.
So a study that nicely assessed the effects of these factors on PPG-derived HR, heart rate and heart, HRV was done by Jessalyn Dunn's group at Duke, and her group evaluated four consumer wearables and two research-grade wearables, and assess the effects of skin tone and activity on the heart rate and heart rate variability accuracy. And what she found in summary on this plot was that basically the device type did matter. that some devices were better for heart rate accuracies. She also found that activity can diminish heart rate accuracy by up to 30% but that skin tone really did not change the accuracy of heart rate and heart rate variability. And I wanted to go a little bit, into the details of this.
So you can see, an example of the effect of motion or motion artifact on PPG. So you can see the motion here. It does affect the PPG signal and when you are going through the signal processing to extract the pulse rate or the beat-to-beat interval, the algorithm mistakenly thinks that there is an additional beat here. And so there is mistakenly assumption that heart rate increases. So you can see that the quality during this time diminishes.
And this is very similar to sort of the confidence interval, the confidence score that is provided to us in, in collaboration with Garmin using our platform. And you can kind of see here, the effects of activity on the quality of the signals.
What we see here is mean deviation error using EKG as a gold standard in various physical activities, including breathing, resting, typing, on a keyboard and walking. And it can be seen here that Apple Watch does quite well, followed by Garmin and Fitbit. And interestingly, the research-grade devices did much worse than anticipated. So the research grades are here demarcated by the light blue section here. And I want you also to be aware though that much of this data may not be, relevant anymore because the Garmin watch that was used during the study was a VS3 or VivoSmart 3, which is several generations removed away from our current VivoSmart 5. And it's also of note is that the Apple Watch is, you know, $400 or when the Garmin watches are about $100 or so. So, I think some of these things, need to be considered as well.
Now this graph, from Jessalyn's study, showed the effects of skin tone on heart rate at both, at rest and activity. The way you assess skin tone is by the Fitzpatrick scale. Basically, it goes from one to six, six being the highest density of the melanin, melanin contact in your skin. And you can see sort of during rest and activity that the skin tone really does not affect the mean absolute error for when you're doing these comparisons. And then when you're actually assessing by device, again, Apple did the best, followed by Fitbit and Garmin, and again, the Empatica and BioEvolution, which are these research-grade wearables, didn't do as well. Then when it comes to heart rate variability, the same sort of conclusion, skin tone does not affect heart rate variability.
Now, what about oxygen saturation. And that's a completely different story. And this was a study that was published in New England Journal in 2020. And I think it received a lot of publicity because it was at the time of COVID. And it was a study done at University of Michigan where they had about 10, 000 measurements obtained from floor and ICU patients and these patients had a paired SpO2, so basically oxygen saturation,derived from a PPG device, an actual blood determined oxygen saturation. So using abg and arterial blood gas. And what they wanted to do was they wanted to assess the rates of occult hypoxemia.
So hypoxemia is defined as an SA O2 of less than 88%, and being occult. If the SpO2 or the oxygen, the PPG device showing what's considered normal. You can see kind of this here in this graph here, the arterial oxygen saturation obtained from the blood and then sort of the 88 percent demarcating whether it's hypoxic or not.
You can see that,in the bottom here is the pulse oximeter, and you can see the anticipated, sort of relationship should be, based on this dashed line here. And you can see the differences between white and black patients, and you can see that the black patients in general had a lower amount of SpO2 compared to the white patients.
So this, you know, I think the conclusion was that there are higher rates of occult hypoxemia in the black patients compared to white patients. But this graph also indicated to me another thing. You could see even in white individuals, white patients here, even, with, you know, a, if a reading of 92 percent using a PPG, there are a number of patients here who actually had low amounts of arterial oxygen levels.
So the surface PPGs may not actually be accurate in both black and white patients in certain patients and there's no way you could predict that. That's some of the note of caution that you have to have here. So I'm nearing the end here. So, you may be asking, so, how does this make any sense?
So, you're telling me that heart rate and heart rate variability is not affected by skin tone, but oxygen saturation is? How does that make sense, since heart rate and heart rate variability uses green light, which we know is sensitive to radiation, to melanin absorption, and SpO2 uses red light, which is theoretically less sensitive to melanin absorption.
Well, it all has to do with the derivation of these measures. So while the green PPG signal is indeed diminished by melanin, the pulse rate is determined by the peak of the PPG signal. So this. this determination is really not that affected by, much by the amplitude of the PPG signal. And so the, the beat to beat intervals, between these two tone colors may actually not be all that different.
For oxygen saturation, however, the melanin in the darker-tone individuals does affect the absorption of the red light more than the infrared light, which then affects the proportion of the red and infrared, signals that then leads to the alterations in the calculated oxygen saturation. And so this is why, you know, skin tone can affect heart rate and heart rate, uh, can affect oxygen saturation, but less so heart rate and heart rate variability.
So here are some final practical recommendations for you when it comes to, obtaining, PPG, PPG derived heart rate and heart rate variability and oxygen saturation. First thing, I really encourage a tight fit. I've seen many patients, who use, watches or study participants using these watches and you can see the light leak out underneath, the watch.
That indicates that the fit is not good enough. I've seen participants sort of complain that the red light is sort of waking them up in the middle of the night. That indicates that the fit is too loose, and that you have to sort of tighten, make it tighter. The other thing that you need to consider is that there has to be adequate pressure of the watch onto the skin.
The ideal sort of level is about 15 to 20 millimeters of mercury. And the reason that has to do with the fact that actually the optimal signal is when the pressure matches the arteriolar or capillary pressures under the skin. It just has to do with the dynamics of that. And so you want to make sure that this is snug enough, but that doesn't really cause a significant skin indentation caused by the watch.
The other thing to consider is that you have to, you should ideally keep the skin warm. I've seen studies, being done in very cold room temperatures, and the PPG signals actually diminishes quite a bit. So, try to keep the skin warm, and if you can't control that, consider using a warming device, maybe like a cloth or a sock around the watch to keep the skin warm.
The other thing, which is pretty obvious, is to minimize movement if possible. also to keep hydrated. And then the other thing that we found is a little bit of a hack is to turn on the pulse ox, at least for the Garmin watches. It seems that when you turn on the pulse ox, which causes the red light to turn on, it also causes the green light intensity to turn up, increase as well.
So turning on the pulse ox may actually increase the quality of the heart rate and heart rate variability measures, but with the understanding that it could significantly drain the battery. as mentioned before.

Course episodes

Sleep 3.1 Introduction to Photoplethysmography (PPG)

Sleep 3.2 PPG Theory and Biophysics

Sleep 3.3 Red vs. Green Light in PPG

Sleep 3.4 Physiological Measures from PPG

Sleep 3.5 Factors Affecting PPG Signals