HRV4Training

Blog post by Marco Altini.
This is the first post on a new blog I'm starting here on the HRV4Training website, where I will write about HRV, hardware/sensors, PPG, data acquired using HRV4Training on me as well as on the growing community of users, and so on. I will try to post data as well as much as possible, so that you can reproduce most of the things blogged here.
Let's get started with the basics. ## What is heart rate variability (HRV)?
Each beat of our heart is triggered by an electrical impulse that can be easily recorded by an electrocardiogram (ECG), one of the most common ways to monitor heart activity. However, our heart doesn't beat at a constant frequency. When we talk about heart rate variability (HRV),
we are interested in capturing the variability that occurs between heart beats.Let's look at 60 seconds of ECG data. This is some data I recorded on myself using the ECG Necklace, a research prototype we developed when I was working at imec, a few years ago. The device is a small sensor connected to 2 ECG leads. Here I haven't converted the values on the $y$ axis to $mV$, but anyways we are more interested in looking at what happens on the $x$ axis (you can click on the image to enlarge):
In technical jargon, the differences between beats are called
RR intervals. The name derives from the fact that the shape of the ECG signal at each beat has been assigned letters (namely the QRS complex). For more on the QRS complex you can just have a quick look on the Wikipedia page, however the only relevant point here is that R represent the peak(s). Going back to my ECG, if we zoom in a bit and look at only 10 seconds, we can see clearly there are differences between intervals; some are shorter, other longer:
Another way we can look at this differences is by plotting an histogram of the RR intervals. Basically we stack up RR intervals that are of similar duration. This way it's actually much easier to see how the values are distributed over a rather wide range. For this plot, I used again the full minute of data from the first plot:
For this minute of data, RR interval values range between 832 and 1094 milliseconds. We can compute the instantaneous heart rate simply as:
$$HR = \frac{60 \times 1000}{RR}$$
The instantaneous heart rate in this case gives us a range between 55 and 72 beats per minute. So far so good, now,
why do we care about these differences in RR intervals?
## Autonomic regulation
The autonomic nervous system (ANS) controls and regulates many functions of our body, from the heart beating to respiration.
The ANS is in control of how our body reacts to stressors, or in other words, the fight or flight response. We typically think of the ANS in the context of its two complementary branches, the sympathetic and parasympathetic nervous systems, that continuously regulate the ANS by acting in different directions. While the sympathetic nervous system is responsible for stimulating the body's fight or flight response, the parasympathetic nervous system is mainly responsible for the body's resting functions.The cardiovascular system is mostly controlled by autonomic regulation through the activity of sympathetic and parasympathetic pathways of the ANS and analysis of HRV permits analysis in this control mechanism [1]. Several studies on HRV highlighted how different features can provide insights in autonomic regulation, sympathetic and parasympathetic activity. In particular, a large body of studies showed a strong link between certain time and frequency domain HRV features, and parasympathetic activity, while findings on sympathetic activity are a bit more inconclusive.Simply put, monitoring parasympathetic activity via HRV can provide insights on physiological stress, with higher level of stress resulting in lower HRV. For example, in the context of sports, heavy training is responsible for shifting the cardiac autonomic balance toward a predominance of the sympathetic over the parasympathetic drive. This means that heavy training will reduce HRV and by monitoring HRV we can possibly optimize training, reduce the risk of overtraining and ultimately improve performance.On these last points, many research studies showed that short-term changes in HRV features, used to assess training load and recovery, are a reliable measure of parasympathetic activity [2, 3] and can even be user to guide training plans [4]. ## How can we determine HRV?
HRV is determined by computing so called features, starting from a series of RR intervals, or differences between heart beats. This means that on the contrary of heart rate, which can be thought of as an almost instantaneous value,
HRV requires a certain amount of data to be accumulated, before it can be computed.Clinical practice recommends 5 minutes of data to be used for features extraction, however in the recent years more and more work was able to show that much shorter windows provide equivalence, and more practical 60 seconds recordings are sufficient [5], especially when we look at time domain features. Another important aspect to take into account is pre-processing to perform on RR intervals before we compute features. One of the most important steps is RR-Intervals correction, which prevents artifacts due to ectopic beats or motion from affecting features computation, as often reported in literature for HRV analysis. It is advised to keep RR interval correction to 20%, meaning that every RR interval which differs by more than 20% from the previous one, will be discarded. Once beats have been discarded, we refer to them as NN intervals, since they are now including only "normal" values. HRV4Training uses a configurable time window, so that you can go up to 5 minutes if you want to, but lets you also take shorter 60 seconds measurements. Additionally, RR interval correction is always performed after the recording, before computing features. ## HRV features
HRV features can be grouped into two main categories, time and frequency domain features. Here is a simple diagram of the procedure that leads to features extraction for both time and frequency domain features:
Frequency domain features require a bit more signal processing, since NN intervals are not at a constant frequency (that is the whole point), and therefore they need to be interpolated before we can do Fourier analysis.
## Time domain features
The most commonly used time domain features are
AVNN (mean of the NN intervals), SDNN (standard deviation of NN intervals), rMSSD (square root of the mean squared difference of successive NN intervals) and pNN50 (number of pairs of successive NN intervals that differ by more than 50 ms). Here is how to compute them given an array of NN intervals of k elements:
AVNN, mean of NN intervals:
\[AVNN = \frac{1}{n} \sum_{k=1}^n NN_k\] SDNN, standard deviation of NN intervals: \[SDNN =\sqrt{\frac{1}{n} \sum_{k=1}^{n} (NN_k - AVNN)^2}\] where AVNN is computed as above. rMSSD, square root of the mean squared difference of successive NN intervals: \[rMSSD = \sqrt{\frac{1}{n-1} \sum_{k=1}^{n-1} (NN_{k+1}-NN_{k})^2}\] pNN50, number of pairs of successive RRs that differ by more than 50 ms The difference between beats is calculated as above: \[NN_{k+1} - NN_{k}\] Then, if $n50$ is the number of beats for which we have a difference greater than 50 ms ($(NN_{k+1}-NN_{k}) > 50$), pNN50 is computed as: \[pNN50 = \frac{n50}{n} 100\] ## Frequency domain features
Frequency domain features can be computed in many different ways, often making it very difficult to compare results between different studies. For transparency, here is how HRV4Training computes HRV frequency domain features:
- collect RR intervals in a time window (I keep all this values in milliseconds)
- remove RR intervals differing more than 20% from the one preceding them (i.e. RR intervals correction)
- interpolate NN intervals at 4Hz (so 1 point every 250 ms, this is necessary since NN intervals are not evenly spaced in time, and we need evenly spaced data in order to perform frequency domain analysis)
- remove the DC component
- at this point I convert into seconds
- compute hamming windowing on the time series I've got from previous steps
- perform FFT
- compute frequency power for the 2 bands of interest (LF and HF).
If any of these steps differs, you'll get different values. Also, sometimes frequency domain features are expressed in terms of relative power and that is again different. I find it hard to directly compare these features to what you can find in literature, but maybe giving the series of steps will help. ## Closing the loop
In the previous sections we talk about how HRV features can be representative of autonomic regulation, and in particular of
parasympathetic activity. We also talked about how certain features are more representative of parasympathetic activity and how a decrease in HRV features values can be indicative of higher physiological stress. HRV4Training reports all features, which can be helpful if you want to run your analysis or are an expert in HRV, but can be confusing at first impact. Many of these features are actually trying to measure the same thing (i.e. parasympathetic activity) and therefore will provide values that are highly correlated over time. According to literature, the most useful features are rMSSD and HF. However, more challenges play a role when using frequency domain features. Tests should be longer, and as I was describing in the previous section, computing frequency domain features seems to be "less standardized". For these reasons it's typically a good idea to look at rMSSD as a marker of parasympathetic activity, rather stable during short term recordings. I will cover other important aspects (breathing rate, time and consistency of the test) in future blog entries.In the next post, I will cover hardware (sensors) that can be used to acquire RR intervals data and alternatives such as using the iPhone's Camera and PPG, a feature that is unique to HRV4Training.## References
[1] Aubert, AndrĂ© E., Bert Seps, and Frank Beckers. "Heart rate variability in athletes." Sports Medicine 33.12 (2003): 889-919.
[2] Garet, Martin, et al. "Individual interdependence between nocturnal ANS activity and performance in swimmers." Medicine and science in sports and exercise 36 (2004): 2112-2118. [3] Pichot, Vincent, et al. "Relation between heart rate variability and training load in middle-distance runners." Medicine and science in sports and exercise 32.10 (2000): 1729-1736. [4] Kiviniemi, Antti M., et al. "Endurance training guided individually by daily heart rate variability measurements." European journal of applied physiology 101.6 (2007): 743-751. [5] Esco, M. R., & Flatt, A. A. (2014). Ultra-Short-Term Heart Rate Variability Indexes at Rest and Post-Exercise in Athletes: Evaluating the Agreement with Accepted Recommendations. Journal of sports science & medicine, 13(3), 535.
## Data
Here are the two files used for this blog post (60 seconds of ECG and RR intervals):
3 Comments
10/7/2015 08:55:01 am
Bravo well written with a limited amount of math that you really dont need to understand the HRV and its different aspects. ALso ( commented elsewhere ) You can tell you are a scientist with research background. While that is sometimes a detriment since scientists can be obtuse, in your case it is an asset. You clearly explained your product and in very easy to follow language. Well a few terms like sigma and mu could be put into english language. Well done. I just bought the polar and downloaded your app. I will let you know in a couple of weeks.
Reply
Marco Altini
10/7/2015 09:33:33 am
Thanks Russ!
Reply
Pierluigi Maini
5/14/2016 01:41:42 pm
Am in full agreement with the comments of my predecessor, rarely have seen a complex topic like this one being explained in such a simple and understandable way: Chapeau!
Reply
Your comment will be posted after it is approved.
## Leave a Reply. |
Blog Index HRV MeasurementsBest Practices Overview1. Context & Time of the Day 2. Duration 3. Paced breathing 4. Orthostatic Test 5. Slides HRV overview Data Analysis1a. Acute Changes in HRV (individual level) 1b. Acute Changes in HRV (population level) 1c. Acute Changes in HRV & measurement consistency 1d. Acute Changes in HRV in endurance and power sports 2a. Interpreting HRV Trends 2b. HRV Baseline Trends & CV 3. Tags & Correlations 4. Ectopic beats & motion artifacts 5. HRV4Training Insights 6. HRV4Training & Sports Science 7. HRV & fitness / training load 8. HRV & performance 9. VO2max models 10. Repeated HRV measurements 11. VO2max and performance 12. HR, HRV and performance 13. Training intensity & performance 14. Publication: VO2max & running performance Camera & Sensors1. ECG vs Polar & Mio Alpha 2a. Camera vs Polar 2b. Camera vs Polar iOS10 2c. iPhone 7+ vs Polar 2d. Comparison of PPG sensors 3. Camera measurement guidelines 4. Validation paper 5. Android camera vs Chest strap App Features1. Features and Recovery Points 2. Daily advice 3. HRV4Training insights 4. Sleep tracking 5. Training load analysis 6a. Integration with Strava 6b. Integration with TrainingPeaks 6c. Integration with SportTracks 6d. Integration with Genetrainer 6e. Integration with Apple Health 7. HRV4T Coach advanced view 8. Acute HRV changes by sport 9. Remote tags in HRV4T Coach 10. VO2max Estimation 11. Acute stressors analysis 12. Training Polarization 13. Custom desirable range / SWC Other1. Intro to HRV 2. HRV normal values 3. HRV by sport 4. HRV, strength & power 5. AngelSensor & HRV 6. HRV 101: How to 7. Top 5 most read articles 8. HRV normalization by HR |

✕