Sleep 7.1 Understanding Cardiopulmonary Coupling in Sleep Studies

This talk aims to illuminate the complexities of CPC, a methodology that has been around for at least two decades but is not widely recognized.

Video Breakdown:

00:00 Intro to Cardiopulmonary Coupling

01:34 Objectives of Today’s Talk

02:40 Traditional Sleep Stage Classifications

09:31 The Legacy of 30-Second Epochs in PSG

12:54 CPC Analysis Explained

14:17 Cyclic Alternating Pattern (CAP) Analysis

This is the seventh talk in our wearable and sleep research series, and today we will be talking about cardiopulmonary coupling methodology, clinical applications, and opportunities. And, you know, cardiopulmonary coupling has been around for 20 years at least. It's not well known to the sleep community, but I would guess that it will likely grow in popularity due to, the increasing use of wearables in the sleep space.
And cardiopulmonary coupling has applications to sleep quality, diagnosing and treating sleep, uh, breathing disorders, so obstructive sleep apnea, central sleep apnea. But the problem is cardiopulmonary coupling is a little bit difficult to understand. So I hope that this talk will dispel some of the challenges of understanding cardiopulmonary coupling.
And hopefully, you'll have some opportunity at the end of this talk to ask some questions that, you know, may come up. So, before we proceed, with the talk, let me just first briefly remind you of our broader webinar series which spans the full spectrum of sleep science, all from a wearable perspective.
The last talk was on sleep regularity and today we will be talking about cardiopulmonary coupling, and next we will talk about circadian physiology and more specifically heart rate variability, temperature, and some additional tests like cortisol and melatonin.
And so what is the objective of today's talk? I hope that through this talk you'll be able to answer these following questions. What is cardiopulmonary coupling and what do its primary measures, HFC, LFC and very low frequency coupling represent? What is the underlying autonomic physiology that links stable sleep to high frequency coupling, sleep instability to low frequency coupling?
How can CPC enhance the assessment of sleep disorder breathing beyond the traditional apnea hypopnea index or AHI? And how is CPC analysis performed and what are the key methodological considerations and limitations of this technique? And what are the principle advantages of using CPC for assessing sleep quality compared to traditional sleep staging?
And I've divided this CPC talk into six categories. Sleep staging classifications, sleep staging through the use of CPC, some of the physiology, how CPC is used for AHI, analytical methodology and strengths and limitations. And so first, a talk about sleep stage classifications. So many of us know the traditional conventional sleep classifications.
The first one was formalized by R&K, back in 1968. There the sleep was divided into wake, REM and non-REM to four stages. And then, the American Academy of Sleep Medicine in 2007 basically kept wake REM, but then consolidated stages three and four into N3. And you could see that sort of demonstrated here again, the four stages previously.
And then in 2007, breaking these two into N3. Now what's happened in the wearable spaces is that the wearable companies have tried to recapitulate these sleep stages according to the conventions that are out there, but they make some few changes and instead of the three stages, N1, N2 and N3, they basically have light and deep stages.
And, you know, based on the discussion, I believe I talked about this, two or three talks ago on the webinar series, and feel free to look it up, is these, wearable device companies use heart rate and heart rate variability in movement in order to categorize into these various stages. Oura is unique in that incorporates respiration and temperature as well.
And the way they do that, and again, I mentioned this in the past talk, is that it's based on the autonomic nervous system, according to the sleep stages. So for non REM as you go deeper and from N1 to N3, you have an increasing, a preponderance of, more influence of the parasympathetic nervous system and less of the sympathetic nervous system. And you could see that more as a stable, cyclical, pattern in, let's just say heart rate. In REM, you're seeing more of a preponderance of sympathetic nervous system. The parasympathetic nervous system is still active, but you can see much more of an unstable disordered pattern.
And so that's what's often used, I would say, by these device companies to help categorize, in the, the data that they're getting from these wearables into these various stages. And so wearables are dividing now, into light and deep, and generally the light stages are basically combining N1 and N2, whereas the deep is correlated with N3.
So, um, these are the conventional sleep classifications. Now I wanted to go a little bit into the visual scoring rules that go into categorizing into these various stages. So the ASM Sleep Scoring Manual was first published in 2007, and there are several things to note about the scoring rules. First, the scoring rules are based purely on subjective visual assessment of the PSG, and it's really up to the sleep tech or the sleep physician to visually distinguish whether a rhythm on an EEG is an alpha rhythm or has a vertex sharp wave.
Whereas sleep spindles has K complex or delta waves. And anyone who's seen a PSG could, recognize it's not always easy to differentiate into these various patterns. And also for the EMGs, you have to see if they have muscle atonia or if there's a reduced tone, in the deeper stages of sleep.
And then for the eye movement, again, it's hard to figure out whether the slow rolling eye movement is different than the rapid eye movement in the EOG. So a lot of these scoring rules are purely based on subjective visual assessments. It really makes it challenging to, do these ratings consistently.
The second thing to note about these scoring rules at these rules were based on expert consensus rather than empirical mechanistic validation. And really they were developed without comprehensive testing of their biological clinical relevance. And sadly, the foundation of sleep fields rests on this visual pattern recognition rather than mechanistic understanding.
So, for example, this definition of the, of slow wave requiring that delta waves occupy more than 20% of the epoch, and that having an amplitude of greater than 75 microvolts is relatively arbitrary. I mean, there's nothing magical necessarily about these 20% cutoff or even the 75 microvolt cutoff because as you age, as we know, the amplitude of the delta wave actually decreases. So, many people in the elderly, stages of their life actually start losing stage N3 and start to be labeled as stage N2 simply because of the amplitude of the 75 Microvolts. This is not only true for the stage N3 classification, it's also true for this, 0.5 second duration of the sleep spindles, the 50% cutoff, for wake versus N1.
And even arousal durations of three to 15 seconds, and then the hypopnea reductions of less than 30% airflow reductions. Those were sort of lucky guesses. I think they did a relatively good job because it seems to stick around, but they are relatively arbitrary and as I've noted, it can change with each individual, so based on their biology. Then the other thing to note is, because of this, because of this is purely a visual subjective assessment, you have moderate to high variability among scores, especially for transitional stages from wake to N1. And N1 to N2. And as we know, the interscore agreement is, assessed by a Cohen's kappa.
And the Cohens Kappa is lowest for stage N1 and for stage transitions, but highest for stage N3 and REM. And the other thing that human visual analyzers do relatively poorly with is with slow changes like for delta wave detections or something called cyclical alternative patterns, which I'll talk about later.
And then the other thing is that human visual analyzers are unable to assess the dynamic relationship across the channels. And this requires complex analytical methodologies, which is, beyond the human visual capabilities or even the analytical capabilities of your mind. And this is a shame, because we are in a digital era of multi-channel recording, and wherein novel, analytical methodologies can hold promising possibilities for better diagnosis in monitoring.
Another, unexplainable allegiance to entrenched practices is our reliance on 30-second epochs, which are universally used in PSGs. This is a legacy from the old analog paper tracing technologies, which is shown here. This is actually from the figure from the first publication done by Loomis in 1930s.
And so he was the, pioneer of using EEGs for assessing sort of the brainwaves during sleep. He used an analog oscillograph. "A" is actually sleep and you could see the cyclical patterns during sleep here. And then you could see this, "B" is during wake. Now, in subsequent decades, sleep monitoring was conducted using ink writing polygraphs, which produced continuous tracings on moving paper.
So a standard paper speed for these polygraphs was about 10 millimeters per second, and so a standard 30 centimeter sheet of paper, which was used came out to be about 30 seconds of data. So a lot of it was just due to convenience because of the speed of the paper and the size of the standard of paper that was used.
And so 30 seconds has been used, and this has been a convention that's formalized in the R&K Sleep scoring manuals back in 1968, and again, repeated in 2007, and remains a practice in modern PSG. It has served its function, but again, this is just purely a, a legacy from old technologies. And so what has happened more recently is that there are newer methods to analyze the PSG more objectively, using repeatable, mechanistic based approach. Probably the more common technique used are spectral based classification. And what spectral, analytical methodologies do is that they break the waves into the frequency components. And one example of this is the EEG time frequency analysis where they're able to break down the EEG signals into its various components.
And this is an example of this, and this is a multi-taper spectral analysis. You could see the traditional hypnogram here. Wake time in pink, purple is non-REM, green is REM and you could see the sort of classic more nREM deep sleep in early stages of the sleep or early parts of the sleep, and then more REM later in in it.
And then you could see, for this multi-taper spectral analysis, some interesting findings. So during wake you see a broad based spectrum activity. When they're close to going to sleep, the person closes the eye, so the alpha rhythm's more prominent. And then as you go into non-REM sleep, you're seeing an increase in the delta activity, the sleep spindles here.
Then during the REM sleep, you start to see the disappearance of this sleep spindles, and then more broad-based activity all across the, the different bands. Then you see some theta bands as well, during the REM sleep. So you can see that this does a very nice job of breaking down sleep into its various stages.
Now, the problem is that this is EEG based and there are different approaches using heart rate and respiration, and this is where cardiopulmonary coupling comes in. Now, cardiopulmonary coupling breaks down the cardiac and respiratory waves into high frequency, low frequency, very low frequency, elevated low frequency, which includes a broadband and narrow band, which I'll talk about later.
But this is how a sleep report from a cardiopulmonary coupling analysis appears like. You could see here that this is sort of stable sleep, unstable sleep, REM, and then wake periods. You could see that during the high frequency coupling, it's associated with stable sleep. Low frequency is associated with unstable sleep.
And the very low frequency is often correlated with either REM or Wake. And so, that is the cardiopulmonary coupling and EEG time frequency analysis, which are spectral based classifications. There's another type of classification, which is based on micro architecture.
And so what I mean by that, these are small patterns which are often missed on the broader categorizations of the 30s epochs that are used in PSGs. Probably the most popular type of classification using this microarchitecture approach is the cyclic alternating pattern or CAP analysis and this is only applied to non-REM.
And what is CAP? So CAP stands for cyclic alternating pattern, and it's an EEG marker of non REM sleep instability. And what this does is reflects the brain's effort to maintain sleep in face of disruptive stimuli. And this stimuli could be external and internal. So external meaning it could be loud noises, or a person being in a stressful environment.
Or it could be internal from, let's say apnea or periodic leg movements. So, these disruptive stimuli can lead to this cyclic alternated patterns and is characterized by two alternating phases. Phase A is the activation pain phase where you have transient arousal vents, which last about two to 60 seconds.
This is a shift towards lighter sleep or brief awakenings based on the EEG. But also you see a surge in sympathetic nervous system activity. So higher heart rate, higher blood pressure, increase in muscle tone, and this phase A is followed by a Phase B or a baseline. This is returned to preceding more stable sleep.
It's characterized by withdrawal of the sympathetic nervous system tone and return to the vagal or parasympathetic nervous system dominance. This returns to the lower heart rate and blood pressure. And this recurring sequence of phase A to Phase B to Phase A to phase B is called a CAP sequence.
Just a little bit, information about Phase A or these sleep arousals. Again, these are EEG defined, so this is an example of sleep arousal, and you could see this activation of the EEG, and then you could see some changes that happens during the respiration, muscle tone as well, the EMG, and also some motion artifact in the EKG.
But also what you see is during this time there's a lag about a few seconds, five seconds of an increase in heart rate, and also a blood pressure that's correlated with this arousal. So again, this arousal is an EEG-defined arousal. It's a physiological paradox because sympathetic nervous activity within a dominant parasympathetic state, the individuals really can't recollect this happening, but it can be physiologically impactful.
And arousals rarely occur in isolation. So it typically occurs in clusters. So this is an example of a CAP sequence, which is unstable non-REM. So you could see this EEG activity, muscle movement, respiratory changes. And then what's on the other hand is non-CAP, which is stable non-REM. Now you may notice an interesting thing.
So both of these are categorized as N2. So even this unstable pattern is N2, whereas the stable is also N2. So, you know, realizing that maybe, these traditional conventional methods of categorization are missing some of these important autonomic and EEG activities.

Course episodes

Sleep 7.1 Understanding Cardiopulmonary Coupling in Sleep Studies

Sleep 7.2 Cardiopulmonary Coupling: Sleep Staging Classifications

Sleep 7.3 High vs. Low Frequency Coupling in Cardiopulmonary Physiology

Sleep 7.4 Cardiopulmonary Coupling and Sleep Apnea Detection

Sleep 7.5 Traditional vs. Modern CPC Methodology

Sleep 7.6 Cardiopulmonary Coupling in Sleep: Strengths, Limitations & Future