Sleep 3.4 Physiological Measures from PPG

• Okay, so for the next one, the next area of topic is physiological measures. 
• Now, this is a typical waveform of the PPG with its characteristic parameters, which includes the systolic peak, the dichrotic notch, and the diastolic peak, and then also the width, at like halfway, at the, between the maximum and the minimum of the PPG waveform.
• Now, there have been extensive efforts to extract as much useful information from this waveform as possible. And, to be honest, it's a continuing evolving field. And one example of this, is, you know, for instance, taking the systolic area. area on the curve for the systolic and also the diastolic portion of the curve. This has been used to assess arterial compliance. 
• Another approach has actually been to use the second derivative of the PPG, basically taking the derivative once and then the second derivative to get this type of waveform. There has been even a significant classification of the morphological features of the second derivative, that can indicate, that whether you're healthy or diseased.
• And so on the left here, you can see, this is a healthy individual and the further you go to the right, the more unhealthy you get. So G, waveform is indicative of an unhealthy sort of waveform.
• Now, for us, however, you know, we're not really focusing on the pulse, PPG waveform. For our intensive purposes, the data that you will be dealing with is primarily focused on the beat-to-beat interval, which is used to derive, sort of the pulse rate or a proxy of the heart rate. So basically we're finding the peak of the PPGs and then finding the interval, the time interval between them to extract heart rate. And as you're extracting this pulse rate interval over multiple times, then you could extract or calculate the heart rate variability. 
• Now, based on the testing that we have done at Labfront, and I think Han-Ping and Francis here are joining this call, are responsible for a lot of this work.
• I think we believe that you can make the safe assumption that the PPG-derived BBI, beat-to-beat interval pretty much closely recapitulates the EKG-derived BBI. And this is an example of this. So the blue is an EKG-derived, beat-to-beat interval, the red is the enhanced beat-to-beat interval that you're getting from the Garmin watch, and then the green is the legacy.
• This is the old, PPG-based, beat-to-beat interval, from Garmin. You could see that the blue and the red pretty much is out pretty significantly. So, I think, it shows that certainly the PPG or the smartwatches can be used pretty well for assessing heart rate and heart rate variability.
• And I think this is particularly during times when you're still, and particularly during sleep, because during sleep, your peripheral vasculature tends to dilate. And so the signals tend to be much better during sleep. Now, there's a caveat here that I want you to be aware of. PPG cannot determine with complete confidence whether the cardiac rhythm is normal, in normal sinus rhythm or not.
• And the reason you can't do that whereas EKG can is because the EKG has something called the P wave here. And the P wave is a sort of a reflection of the normal SA node conduction along the atrium. You have the electrical conduction from the SA node to the, through the atrium to the AV node and it generates the P wave here.
• So the existence of the P wave indicates a normal rhythm. The PPG, on the other hand, has no such marker. There is no such P wave. And so basically there really is no great or confident way that we could assess whether the rhythm is truly a normal sinus rhythm or not. If there's a significant amount of interbeat variability, then, you know, obviously with atrial fibrillation, there may be such variability that you could be able to detect that.
• And certainly Apple and other companies have worked on doing that. 
• The other important thing to note is that there is usually a time lag of just under 300 milliseconds from the peak that you identify the EKG to the peak to the, that you get from the PPGs. And this time lag has to do a number of factors.
• First, the electrical activation is the one that generates this signal and this peak wave. Then, you have this delay due to the ventricular mechanical, sort of electromechanical delay. Also, the pulse translate time, basically the time it takes for the pulse pressure to go from the heart to the watch or to the wrist, and then from the deeper layers to the superficial capillary.
• So this, you know, ultimately adds up to about a 300 millisecond delay. And this delay is sometimes really helpful to kind of extract, or derive the blood pressure. 
• There's a time lag between the R wave of the EKG and the peak of the PPG, and that time lag is called the pulse arrival time.
• And this value is actually a function of blood pressure. And so the higher blood pressure causes the arterial, arteries to distend, and the arterial wall tension to increase, and then subsequent leads to a higher elastic modulus, or the Young's modulus, which then leads to a higher, a faster velocity of the pulse pressure.
• And so the higher pressures that you have, blood pressures you have, the faster the blood vessel, pulse pressure travel, and the shorter time that it takes from the peak of the EKG to the watch. And so this is how, you can extract blood pressure. We're using the combination of the EKG and a wrist watch.
• The PPG can also be used to derive respirations, and this is done by four ways. The first is, you know, this is, you know, assuming that there's no modulation of the PPG, but with respiration, you can get a baseline wander here. Basically sort of like this slower variation of the intensity of the PPG. And this actually is more a reflection of the venous blood volume under the skin because the venous blood volume is much more sensitive to respiratory mechanics.
• The other thing that you can get is sort of the frequency modulation. So, as you may know, is that when you breathe in, your heart rate increases. As you breathe out, your heart rate decreases. And so you could use that in a way, to sort of extract respirations by assessing the frequency modulation of the pulse rates that you see in the PPGs.
• There's also amplitude modulation. As you breathe in, actually your blood pressure drops. I don't want to go into too much of the details as to why that happens, but as you inspire, the amplitude of the PPG decreases, and as you expire, the amplitude increases. And so you could use the amplitude modulation as a way, to extract respiration.
• There's also pulse width modulation. You could see that it gets wider and thinner, during each of these processes. So there's altogether four different ways of extracting respiration from PPG. 
• So what is the effect of the measurement site on the PPG waveform? Now, this was a study that assessed the amplitude of PPG across multiple sites, and it was done at the finger, the wrist, on both the dorsal and the volar aspect of the arm, arm, earlobes, and the fore, forehead.
• And you can see at least the finger has the greatest intensity or amplitude of the PPG, and the wrist, particularly at the dorsum, where the watch is often located, has a very diminished amplitude. You can see that in both sort of normal breathing and deep breathing. But I just want you to note that this study used a reflectance near-infrared light PPG, not the green or blue light PPG.
• So you can't necessarily, you know, extrapolate this to the PPG signals that you get in the watches. But I wanted to also show you sort of the normalized, sort of the waveform differences between these different site locations. And this is the normalized average of the PPG across the sites. And you can see that the finger peak is sharper, while the wrist and forehead, it's a little hard to see here, but the forehead I think is in gray and the wrist in green, tends to be sort of wider and broader. 
• The difference between these locations and the appearance of the waveforms can be attributed to the finger being much more sort of vascular. There's more densely packed arterioles and capillaries in the fingers. They also have arteriovenous anastomosis.
• Because it's predominantly arterial at the surface and less venous. And this is important because, at least in the wrist, you have much more of a venous component. And if you can recall, in the previous slide, the variation in the venous pulses causes that wandering baseline that's associated with respiration.
• So actually the wrist location may actually be better to detect respiratory rate than the finger is because the finger doesn't have that venous component as much. The other component is that the rich, um, that the, the fingers, the vasculature has a very rich, innervation of alpha anergic, nerves and receptors. And so it can be sensitive to sympathetic nerve sense, since system activity and also temperature as well. 
• There's one additional feature of the PPG waveform that I wanted to sort of evaluate, sort of discuss about, and that's the diacrotic notch in the diastolic peak. Now, the diacrotic notch is associated with a closure of the aortic valve.
• Basically, it's the retrograde flow in the aorta associated with a relaxation of the left ventricle. The diastolic peak has to do with a pressure reflection from the lower extremity. So, If you can imagine, you have a contraction of the heart, which generates a pulse wave that goes both to the arm and the lower extremity.
• Now, the pulse wave goes into the lower extremity, and then there's a reflection that goes back up. And this reflection is due to some impedance changes that you get either at the bifurcations or the smaller vessels. And that reflection then subsequently shows up in the wrist.
• They use this as a way to assess for arterial compliance. And this is an example of this. So you could see here, for the younger individual, the time it takes for this reflection from the lower extremity to the upper extremity is longer because, you know, the arteries tend to be more compliant.
• However, as you age, your arteries get stiffer, less compliant, and so the velocity, um, that velocity, uh, for the pulse increase, is quite, um, increased, and so leads to a rapid reflection and earlier appearance of this diastolic peak, um, in elderly individuals. And so you could also sort of understand why the diastolic peak in the finger is a little further on or later, compared to the wrist. It's just because the time it takes for that reflection to go, to the wrist happens earlier compared to, to the finger.
• There are also some PPG changes that are seen across the sleep stages too.I don't want to go into too much of this detail because that's going to be a topic for the next talk, but you could see here that during wake and REM, that a lot of the PPG signals are, are pretty consistent and the amplitude is large. But as you go into deeper sleep, you have, smaller amplitudes and also greater variations in the PPG signals, across the sleep stages.
• A lot of this detail or this information, unfortunately, is not available in the Garmin watches that are used for Labfront, but we are looking at some point to hopefully extract raw PPG, in some way or form, either in collaboration with Garmin or incorporating an additional device if needed.

‍