3.5 Other Nonlinear HRV Methods: MSE, Symbolic Dynamics, and Poincare
This video covers multi-scale entropy, symbolic dynamics, Poincare Plot and Dr. Ahn's thoughts on non-linear HRV and HRV as body-wide function.
- 0:00 - 3:15 Multi-scale Entropy
- 3:16 - 4:09 Symbolic Dynamics
- 4:10 - 6:00 Poincare Plot
- 6:01 - 10:07 Personal Thoughts on Non-linear HRV
- 10:08 - 11:19 HRV as Body-Wide Function
This video concludes HRV Course: Part 3. Visit Labfront Academy for more resources!
- Finally, I like to summarize other nonlinear Heart Rate Variability measures. Here's a limited time, I'm not going to go into all the nonlinear other alternative nonlinear measures in great detail. But the one that has probably received the most attention is multiscale. Entropy multiscale.
- Entropy assesses the regularity of the signals, and it's done over multiple scales. And how regularities is assessed by is by using something called Sample entropy. And sample entropy assesses the relationship sort of the consecutive ordering of the data. So in this case, you take a look at these first two points and you identify in the subsequent datasets later on, is to see like where do you see similar relationships occur. And so you can see that you have one occur here, and another one occur here. But then you could also see for the third point in this relationship, you see something that's blue here. And so you try to see again, do you see the same subsequent ordering in other identical sets, and you don't see that here, and so this one's not included. But this one certainly has that third consecutive point which matches this. And so this sample entropy is essentially out of the two possibilities of similar options, you have one that appears similar like this. So the lower number of similarities that you have compared to the available options will lead to a higher overall sample entropy. So a higher sample entropy is attributed to greater irregularity of the signal. And you do the same approach. But you do this across multiple scales again, similar to the DFA a scaling as well.
- What you get is graphs such as this, and this is an example of one which was taken from fiscal reviews letter in 2002. You have healthy individuals, which is represented by this green graph, atrial fibrillation in this graph, and then congestive heart failure in the circle graph here. So atrial fibrillation individual tends to have a lot of the irregularity occur in the high frequencies, but then as you sort of add go out into the lower frequencies, that you you start losing that irregularity. However, for congestive heart failure, us tend to have low amounts of regularity across all temporal scales, in both low and high frequencies. And the healthy individuals have maintains high amounts of regularity across all temporal scales. This could be extended to differentiating between the young and elderly, young individuals have an overall greater irregularity or a sample entropy across all the timescales compared to the L theory. So if you take a sum of all these values across all the scales, the some ms E or the multiscale, entropy tend to be larger in in young individuals compared to the elderly.
- Another nonlinear measure methodology is this symbolic dynamics, symbolic dynamics, just like NSE does care about the consecutive ordering of the our interval data, but they use symbols in order to describe sort of the patterns that we see here. And you could see, for instance, there are different patterns here, some that has very low, very guilty, one that sort of maintain some consistency, but then later, some variation, and then one that has variations in both consecutive measurements. And what this study is published in 2005 by Cassetti, show that those individuals with higher numbers of zero variants tend to be more prone to arrhythmias compared to those who did not get arrhythmias.
- That's the type of relationship or type of methodology is the point Chari plot analysis. What this does is you map out what are integral compared it to the subsequent our interval. And from this graph, you get this these plots and then you try to extract the elliptical fit for that, and then you extract SD one, which is the short axis for this ellipse and the SD SD two is the long axis of the ellipse.
- From my perspective, Porcari plot is most useful for detecting noise or ectopic beats. And this is an example of why you could detect these ectopic beats. You normally have these normal beats which have usual our interval but then when you have an ectopic atrial beat, you have a short r followed by a long R interval. So as a result of that this short R may be mapped up in this region here, and then it's followed by a larger R interval. And so you could see these these possible ectopic beats. And then certainly this large one is followed by an irregular one, so you could get another ectopic beats or another plot in this region as well. In truth SD one in st two can be derived by using just simple time domain measures. SD one squared, for instance, is equal to one to one half of standard deviation of standard deviation squared, this is a time domain measure. SD two can also similarly be derived using just time domain measures itself. So you do not necessarily have to plot these graphs out in order to get SD one or SD two.
- Here are my personal thoughts on nonlinear heart rate variability, enthusiasm in truth has faded over recent years. And there are a number of factors for that. First, you need long term recordings. And if you remember, for power loss, one a ref slope and specifically, you're required at least 24 hours of data. And that also remains true of DFA alpha one. And the other reasons why leucine has faded is that a lot of this study has been limited to centers with advanced knowledge and engineering computer science. In fact, most of the studies that were done in the nature studies done for DFA Alpha One were done in Finland, particularly by the university oluwo and medical center there. So you know, there are very few other centers that did look into this in great detail, but not too many. The other limiting factors, it's quite sensitive to adjustable parameters and pre processing and de noising steps. That's something that needs to be elaborated further. There's also limited info about reproducibility, probably due to some of these adjustable parameters, but also due to noise from the signals itself. And probably the biggest problem in regards to nonlinear heart rate variability, is it difficult to understand from a physiological standpoint, what is attributing to a Brownian noise, or a white noise signal, for instance, is something that we have not fully established yet. And for this reason, it's unclear how clinical care is going to be changed, but what can we do to improve it? If we really don't understand the physiological basis for it, then it makes it hard for us to figure out what to target in order to treat it accordingly.
- But from my opinion, there's clearly something there. We've seen that nonlinear Heart Rate Variability measures have often outperformed the traditional Heart Rate Variability measures with regards to mortality and number of situations and conditions. And we're seeing that, you know, nonlinear measures are being increasingly available and accessible to ordinary citizens, but absent data availability. And, you know, I think the intra individual day to day variations will likely you personal specific information. As I mentioned, this is an area that has been under studied, the dynamic changes that we see in heart rate variability, not only if the traditional but also the fractal related measures have not really been evaluated in great detail. And just like the application that we saw for lactate threshold for exercise, I think the same can be extended to many other conditions. And it's an understudied area. And with the availability of wearables and apps, and etc, there's going to be I would predict much more activity in this area.
- And DFA alpha and these fractal scales and nonlinear measures can potentially be insights into energy allocation distribution across organ systems. The theory of behind a lot of these measures is that you have a heart that is sensitive to systems that operate across multiple timescales. So you know, for respiration that operates in certain scales, higher scales, you have better reflects the operates and lower frequencies, then you have thermal regulation, and Renan angiotensin switch operates at the very low frequencies, etc.
- So if you have a heart that functions well, it operates all across all the scales, and the nonlinear measures should do a particularly good job of assessing all across those scales. And I think as we pay more and more attention to positive health and not just purely on disease, some of these nonlinear measures which you know, track much better to positive health overall, will receive much more attention. So I think there's a lot of excitement in nonlinear heart rate variability, and it's something that I think we should pay attention to in the future.
- Hopefully you could see why I presume the heart rate variability can be used as a marker of body weight function. First, it has been applied ecological in ecologically consistent environments. It's a change in usual living, free running conditions. There's no need for controlled settings.
- The other thing is that there is no single temporal scales, the heart interacts across different organs with different timescales. And we've seen this in particularly with a fractal measures. And the heart rate variability measures certainly focus on dynamic patterns that emerge across multiple timescales. So comparing sin to the 1980s and 1990s, where the focus was primarily on autonomic nervous systems, the heart rate variability started to focus now on the human body overall, and its function in usual day to day living. So I think from this perspective, this is why I think heart rate variability is serves as a marker of body weight function.