How Does HRV Work Under The Hood?

These are my notes from a recently webinar by Dr. Andrew Ahn, an Assistant Professor of Radiology at Harvard Medical School whose research interest is in computational medicine and in large-scale electrophysiological process within the human body. He is an unpaid scientific advisor to PhysioQ, which makes measurement tech for research.

Dr. Ahn’s lecture was wonderfully clear, and I found myself taking detailed Roam notes. I’m sharing these notes below. They were taken while watching the video, with just a few later corrections, so please forgive any errors or mistakes. I have an intuition that they will be of interest to at least a few people here! This is part one of a lecture series by Dr. Ahn. Part two is being given September 13th and I’ll try to watch. If you want to watch it yourself, here are the details:

Notes from Heart Rate Variability: Antiquated or Indispensable? Part 1

  • Author::Andrew Ahn
  • Publication Date:: August 2021
  • URL or Bibtex:: HRV: Antiquated or Indispensable? Part 1
  • [[Reference Notes]]
    • HRV is considered antiquated by clinicians, and useful by athletes and others involved in self-tracing. Interestingly, many clinicians do not know how highly valued HRV is outside the clinical setting, and many non-clinicians using HRV do not know how skeptical clinicians are of it’s utility.
    • The Zio continuous HR monitor, which is commonly used in clinical practice, does not even report HRV.
    • The last time any major guidelines or consensus paper was published was 1996. European Society of Cardiology and the North American Society of Pacing and Electrophysiology, published in European Heart Journal and Circulation. They only identified two uses: post heart attack assessment, and diagnosis of diabetic neuropathy. The only follow up was in 2015, a position paper by the some of the same group. They looked at non-linear method and found no reason to recommend them over traditional methods. It was published in a much smaller journal and wasn’t very much cited.
    • Wearable companies use HRV as a cornerstone of app functionality. Whoop Recovery measure uses HRV as one of the core measures.
    • It has also become more interesting to academic researchers, even when rise of # of journals is taken into account.
      Increase in HRV publications
      • Obstetrics was the most prominent discipline in the 1980s, because of their interest in fetal heart rate monitoring.
      • Then cardiology dominated 1990s and 2000s.
      • Then neuroscience started to take over. Autonomic issues were the first, and then the development of neuroimaging, engineering, and sports medicine.
    • Stages: Understanding HRV, HRV as a marker of autonomic nervous system, HRV as a marker of body wide function, HRV within a construct of mind-body interaction, HRV as itself a desirable target.
    • A very good physiological description of the nervous system influence on heartbeat. I learned that the “natural” heart rate (that is a heart rate that isn’t connected to the nervous system) is rather high. They know this from humans and animals who have had heart transplants. I also liked looking at the basic schematic of influences on HRV, including central neural control and peripheral mechanical factors.
    • In the 1970’s the identified the phenomenon of respiratory sinus arrhythmia, which is the rise and fall of HRV as you breath. Inhaling quickens the heart rate. Exhaling decreases the heart rate. (I learned this in yoga class a long time ago.)
    • The Baroreflex is a nervous system feedback loop linking blood pressure with heart rate.
    • There are a number of different ways to measure HRV
      • Time domain measures
      • Frequency domain measures
        • Time series analytic techniques such as Fourier Transform are used to evaluate the power of each frequency range. (We start to get over my head here, though I’ve benefited from some help doing similar analysis on my own tremor data.)
        • The frequency categorization of HRV described by Dr. Ahn involves the following bands:
            1. High Frequency (~.35Hz) associated with respiratory sinus arrhythima
            1. Low Frequency (~.1 Hz)associated with baroreflex
            1. Very Low Frequency (~.025 Hz)associated with temperature regulation and certain hormonal influence, including reproductive hormones and steroids
          • 4 Ultra Low Frequency associated with thermoregulation, other hormones, including cortisol, and circadian rhythms
        • (I do not know what the difference is here between “temperature regulation” and “thermoregulation.” This could be a mistake on the slide or a real distinction I missed.)
        • The low and high frequency cycles can be evaluated within a few minutes. The variable low frequency and ultra low frequency require 24 hour measurement.
        • High frequency range (.15-4Hz )is associated with the parasympathetic system.
      • Non-linear measures
    • Dr. Ahn then goes to the cellular level and talks about the pacemaker cells in the sinoatrial (SA) node. If you would like to follow at this level of detail, definitely watch the video. His explanation is very clear. (It begins at about minute 28:30.) The upshot, however, is that only where there is normal sinus rhythm, with heartbeats originating in the SA node, will HRV provide insight into parasympathetic activity. Therefore Dr. Ahn describes the correct beat-to-beat interval to measure as the “NN interval,” rather than the more commonly seen “RR-interval”.
    • Therefore, HRV, which comes from an analysis of the time between heartbeats, has to be evaluated in the context of other data that can indicate whether there is a normal rhythm or a period of abnormal rhythm, such as pre-atrial contractions or atrial fibrillation. This data is available from an EKG.
    • But what about when measurement of HRV is done using a smartwatch? Smartwatches give us HRV from Photoplethysmography (PPG). They measure the flow of blood through your skin. You can’t directly tell, using traditional methods, whether and what kind of rhythm you have. A lot of abnormal beats will give you a high HRV reading with these systems, which does not reflect autonomic activation. (This part of the talk was very interesting to me.)
    • Sympathetic and parasympathetic influence on HRV have different dynamics, as might be expected since their effects are seen in different frequency ranges. The high frequency, parasympathetic-associated cycles have quick onset and offset, and respond immediately to breathing. The low frequency, baroreflex-associated cycles have a relatively delayed onset and then a slower recovery. (Dr. Ahn goes over these dyanmics in detail around minute 42:00.)
    • So what causes the respiratory rhythm in heart rate? It is because inhalation reduces the activity in the vagus nerve, which is the parasympathetic inhibitor of the “naturally” high heart rate.
      • AMAZING DETAIL in this section: Dr. Ahn expresses surprise that the researchers who proved this conjecture were able to find volunteers for their experiment, which involved sticking a needle into the neck of the subject, near the carotid artery, and directly recording vegus nerve activity. He speculates that the investigators performed this research on themselves.
      • Here is a video of one of them describing their recordings. He does not mention this detail and I have no idea if it’s true.
        YouTube
      • Here is a link to the published paper: In vivo recordings from the human vagus nerve using ultrasound-guided microneurography
    • It turns out there is also respiratory influence on the sympathetic nervous system. (This is easier to measure since you can get your instruments into a peroneal nerve in the leg without worrying about pricking your carotid artery.) You can also pick up these systems using electrodermal skin conductance. This works because the sweat glands are activated by the sympathetic nervous system, and sweat will increase skin conductance. However, these show up in heart rate as lower frequency cycles.
    • If you can increase your HRV by breathing slowly, what happens if you breath very slowly, more slowly that about seven breaths per minute? Won’t that change the dynamics and possibly mess up the distinction between high frequency parasympathetic-associated high frequencies and the lower frequencies? The answer is yes, and some details are explained at minute 47:00.
    • The result of these interactions is what is described in Dr. Ahn’s talk as “the coherent state.”
      • Interestingly for me, he uses a figure created by HeartMath. Some of the earliest talks about HRV at QS meetings discussed the Emwave device from HeartMath, and the Emwave was the first instrument I saw and used to output an HRV-based measurement. (I don’t think it gave an actual number; I seem to remember red and green lights, though I may have this mixed up) I had lost track of them until I saw this slide.
    • The coherent state, according to Dr. Ahn’s report of current research, occurs at .1Hz, or 10 seconds per breath. Dr. Ahn mentions a popular belief that this coherent state “retrains” your autonomic nervous system in a way that is beneficial to health. He finds this interesting, but he does not have research to support this idea.
    • Now that we know we are measuring “NN-interval” and why, Dr. Ahn’s talk turns to the analytical details. There are four main HRV values in use:
      • Standard Deviation of NN-Interval (SDNN)
        • Risk of mortality in cardiac patients
      • Standard Deviation of 5-minute averages of NN-Interval (SDANN)
        • Predominant circadian rhythm
      • The Root Mean Square of Successive NN-Interval Differences (rMSSD)
        • Vagal modulation of heart rate.
      • The Percentage of NN-Intervals that Are Greater than 50ms Different than Prior NN-Interval (pNN50)
        • Sensitive to uneven beat detection.
      • I can imagine being interested in knowing how a device calculates HRV, but I do not do any work myself related to selected these calculations, so I didn’t bother noting more details, but Dr. Ahn begins discussing them at about minute 54:00, and even provides a glimpse of the calculations involved — probably enough to follow up on if interested.)

That’s it for now

I will probably take notes on part 2 of this lecture and post them when done. Here’s a link to the next event, September 13, if you want to view it yourself.
Heart Rate Variability: Antiquated or Indispensable? Part 2

3 Likes

Here’s the jumping off point for an excellent guide from @Marco_Altini with practical best practices for getting a valid HRV measurement:

I’m also interested in HRV and tracking it for about a year. Here is my findings

  1. I’ve quantified my physical activity and found a strong correlation with my HRV
  2. I’ve tracked my illnesses and also found strong correlation with my HRV
    Also i’ve found a better sensivity of standing measures (Polar H10) compared to nightly average (Oura ring)

@Max_Eastwood have you experimented with deliberate influence on HRV (for instance, using slow breathing) and found a range for these effects?

Thanks Gary for posting this. Great summary.

Parts 1 & 2 talks (out of anticipated 5 total Parts) on HRV are available on YouTube, if you wish to watch in your spare time.

Part 1: Physiology & Methods Heart Rate Variability: Antiquated or Indispensable? - YouTube

Part 2: Clinical Use & Factors that Influence HRV Heart Rate Variability: Part 2 - YouTube

Parts 3-5 are in the works and will address HRV epidemiology, non-linear approaches, neuroscience underlying HRV, and applications in wearable science.

Let me know if you have any questions.

@Andrew_C_Ahn : After I watched the first part of the series, which is excellent, I saw this paper:

Comparison of Heart-Rate-Variability Recording With Smartphone Photoplethysmography, Polar H7 Chest Strap, and Electrocardiography by Daniel J. Plews, Ben Scott, Marco Altini,Matt Wood, Andrew E. Kilding, and Paul B. Laursen

The abstract (I don’t have access to the full text) says: “Both PPG and heart-rate sensors provide an acceptable agreement for the measurement of rMSSD when compared with ECG. Smartphone PPG technology may be a preferred method of HRV data collection for athletes due to its practicality and ease of use in the field.”

What do you think? My understanding of @Marco_Altini’s point of view is that you can compensate lack of precision in optical measurement by controlling the measurement conditions.

I think that PPG is definitely a viable alternative to ECG for calculating HRV - particularly when the wristband is fit snugly (not overly tight) and the person is at rest.
However, there are several caveats:

  • PPG cannot truly detect if a person has arrhythmia (particularly well-controlled atrial fibrillation).
  • PPGs don’t work all that great during rapid movement (exercise d/t noise) or even isometric exercise of the upper extremity (d/t compression of the microvasculature)
  • Finding the exact point to define the timing of each pulse (from PPG) is not always straightforward: timing of the PPG peaks can depend on various factors [many companies actually use the initial rise of the pulse as a marker - or so I have heard]
  • Use of PPG assumes that the pulse transit time is always constant. This is always not the case since conditions such as large pericardial effusion or asthma exacerbation - can theoretically alter the pulse transit time w.r.t. respiration.

But as a general rule, PPGs are pretty good IMO. RMSSD also are reported to be less sensitive to the occasional noise/ectopic beats…

In my experience, the signals from the Polar H7/H10 are really good. Pleasantly surprised with the quality…

Hope this addresses your question?

  • Andrew

@Andrew_C_Ahn Yes, thank you! I’m experimenting with HRV4Training now, which uses the camera. This would seem to be a problematic way to measure HRV, especially since I have frequent arrythmias (PVCs), on the worst days >5% of heartbeats.

However, the measurement protocol involves short measurements at the same time each day, under controlled conditions. I’ve been tracking my arrhythmias for over a year, with more than 10,000 observations, and a side effect is that I’m pretty aware of when they are occuring. Also, they show up on the real time visualization when they occur. So I’m going to see what I learn from measuring in this way for a few weeks every morning.

@Andrew_C_Ahn What is your opinion of the Polyvagal Theory by Stephen Porges and can HRV biofeedback be successfully used to overcome trauma as he defines it?

Polar H10 measure HR by electrodes, bot by PPG.

That’s seems to be true, but not for a shot-term measurement. For example, oura ring HR / HRV was scientifically validated and showed that nightly HR / HRV are pretty accurate, but 5 minute measurements not. Since i use both oura and fitbit, which also measures HRV i can compare them.


As we can see, nightly HRV trend seems to correlate well, but a few night they differs.


Here is correlation coefficient 95% confidence interval [0.87-0.94] (bootstrapped by BCa). CI is narrow and confirms large correlation.
But whats correlation for a 5-min periods? It’s [0.63,0.63] and we can see on a 5-mins intervals devices agree less. So i take with a lot of salt short term measurements with wrist / finger ppg and relay on average long term rest period (sleep at night).

Also it’s worth to note that HRV may be affected by some medications or atrioventricular block. Atrioventricular block will influence HRV in way that it becomes totally useless (not reflecting ANS activity because HR coming not from sinus origin, which we try to measure and interpret). You can read great topic on researchgate especially Przemyslaw Guzik comments.

Also i’ve found that my breathing are pretty slow, around ~11-12 breath per minute, which moves my RSA peak from HF to LF making them not interpretable. After i’ve realized this, i’ve started looking only at RMSSD.

Finally, i’ve found the shortest way to measure my hrv and look trends: morning 2-minute standing (1 min stabilization period) with Polar H10. In my case standing HRV was more sensitive to physical activity and sickness compared to average nightly hrv.

As i said above, i’m already a slow breather with 11-12 breath per minute. When i measure my standing hrv, i just trying not to influence my breathing and have regular rhythm. I’ve read 100+ papers about HRV for last year and there is only a few about paced slow breathing. I remember 1 or 2 which not recommend to influence breathing if you want to measure HRV.

Gary, I too had multiple PVC’s in my 30-40’s (age that is) and, for some reason, it has gone away. I wish I knew why, but I presume it was because I don’t push myself too much when exercising/running now and I just basically listen to my body rather than my mind. Sounds corny, but it is what it is. :slight_smile:

One option for these PVCs (if they are not consecutive) is to eliminate those pulses which clearly create an abnormal pulse-to-pulse (PP) set of intervals. PVC/PAC typically cause a short PP-interval followed by a long-PP interval. You can then place in a new pulse that is interpolated based on the adjacent PP-intervals. It’s a bit laborious, but a consideration if you really wanted to know your underlying HRV.

I guess one advantage of having PVCs is that you can theoretically calculate HR turbulence (from continuous ECG) to get a sense of your baroreflex function. But that too is a bit laborious…

@jntamm I am actually a fan of Polyvagal Theory and find it quite compelling… It has provided me with a larger framework to understand vagal function and has helped me resolve some of the contradicting aspects of vagal activity seen in the clinical setting (being bad for neurosyncope, but good in preventing arrhythmias, for example). I was planning to talk about it into Part IV of my talk series.

I wish that I could tell you more about Biofeedback in its role of address TBI, trauma, etc… Haven’t had a chance to review the literature yet, but hope to do so in the future. In general though, I am more optimistic about its potential benefits than many of my clinical colleagues.

@Max_Eastwood Oh, I am aware that Polar H10 is an ECG. I think I mentioned it because it was mentioned earlier in the thread.

I agree with much of what you say here. I don’t have much experience with Oura or Fitbit - so this is a good time as any to learn from you how they produce their results. I generally analyze raw PPG data from medical-grade devices, physiological datasets and so my conclusion was based off my experiences from that. However, I presume that each of these companies have their proprietary approaches to producing HRV measures (how to detect pulse time, prepocess the data, apply frequency analyses, etc…) so we may never know whether the lack of consistency in 5-mins period is due to differences in data processing/analytical approaches or something intrinsically faulty about PPG.

Yes, agree that AV block (2nd/3rd degree, but not 1st degree) can make HRV data useless. We also similarly found in Taichi masters who generally breathed < 9 bpm that their RSA entered into the LF range and thus made frequency domain HRV difficult to interpret (I address this slow breath effect briefly in Talk 1).

Your sets of experiments on your blog are quite impressive. Thanks for making it available for everyone to see.

1 Like

Switching gears for one moment here. Anybody else curious about the Matteo M Ottaviani, Leah Wright, Tye Dawood, Vaughan G Macefield self-experiments (if that’s what they were)? With just a hint of encouragement I’m probably vulnerable to going down this rabbit hole and trying to see about an interview: will they admit to being their own research subjects?

I hope this is not too off topic:

I was hoping to use HRV as a biomarker of aging (at least of the cardiovascular system), but I have
came across following article that puts a doubt on the applicability of HRV for this purpose:

Given the size of the sample, and length of the followup in the study, I find it quite compelling, but
would anybody care to comment on it if I missed something?

I agree - this is too simplistic, but HRV does comment on cardiovascular health but is only one of many factors.