Recently a study came out examining the accuracy of several optical based monitors and the Polar H7 chest strap compared to EKG tracings.
This was done in a lab setting but did include treadmills, bike, and elliptical machines. The results indicated the chest strap was closest to EKG accuracy with various optical devices further down the list.
Interestingly, the Apple watch performed the best of the optical group!
In any case the scatter was relatively high in the optical group indicating less than ideal accuracy.
Why the need for accurate peak heart rates during intensity?
Generally we are not very concerned with some mild deviation at low or moderate intensities where the typical wrist units are fairly accurate. However, at higher intensity there are several potential reasons depending on the particular issue of concern. For example, to reach near VO2 max, one needs to attain a near max heart rate. If the device can't accurately track this (given the challenging circumstances), it would be difficult to know if VO2 max was reached.
Another scenario is the differentiation between over reaching vs de-conditioning. Let's say you have a below average power (or running time) during an intense interval of known power/heart rate. The question is why? If you had taken 2 weeks off and were de-conditioned, one could see a normal max heart rate but lower than normal power. With over reaching the power would also be reduced, but the heart rate would be less than usual:
The over reached individuals had a decline in power and heart rate. The heart rate numbers were just a few beats off, indicating the importance of an accurate, trustworthy assessment. So knowing peak heart rate vs power can help guide whether you are over reached or just need to train harder!
The fact that optical wrist based monitoring is not very precise at high intensity is probably recognized to anyone who has ever used them. Unfortunately, if the results are not precise enough, important training and physiologic status insights could be wrong.
Are belt type units truly the "gold standard"?
One may conclude the the belt type monitor should be equivilant to an EKG derived heart rate given how closely it tracks to the EKG.
However, lets look at this assumption more carefully. In laboratory medicine, it is very important to have controls and calibrations when using your equipment. For instance, your glucose monitor may give a numerical result, but it may be erroneous. This could be related to old glucose strips, low battery, or dirty sensors in the meter as examples. Coming back to belt type monitors, do we have any feedback data indicating that the tracing is valid? The answer is no. One may say that a visible tracing is not needed but the following examples will show why it is important. Some time back I noted that the Hexoskin was not reading my heart rate properly. Fortunately, the Hexoskin web site (as well as the smartphone app) can show you the raw data. The following three tracings of raw data show the problem:
The above basically shows a bunch of random noise. The output module will transmit a numeric heart rate but obviously it is worthless. The cause for this was a combination of a cool day, no sweating and no conductive cream.
Some data in noise:
Here there is some information in the noisy signal but the numerical output would be less than trustworthy. Causes are possibly the same as above
This was taken from a very high intensity interval with a derived heart rate of 170. It is a well represented 1 lead EKG and the user can be assured that the number is accurate.
My procedure now is to look at the smartphone tracing before I start riding to make sure it does not look like this:
If it does, I apply more conductive cream and make sure there is a clean signal before starting a session.
Even "gold standard" devices can lose precision:
Theoretically a belt strap should have similar accuracy to an EKG, but since there is no raw data to confirm this (at the time of measurement in the field) it is not a certainty. Sure, in a lab setting, they would have excellent correlation, but let's say you were interested in monitoring your max heart rate. How would you know that the signal quality (and numerical value) were associated with a good signal at the same moment as you reached maximum?
The fact is, belt straps at their best can rival EKG accuracy but we have no guarantee of this at the time of measurement.
Optical heart rate done right:
Another observation is that wrist/arm based monitoring is near worthless during a session of very intense exercise. A previous post looking at muscle O2 saturation in upper extremities indicated very poor tissue perfusion at high intensity. I surmised that this (plus high extremity motion) would make it difficult to accurately measure heart rate by the optical method if the sensor was on the arm.
It was also discussed that moving the sensor to a better perfused, more stable location should be optimal. I recently came across the Moov Sweat heart rate device which is specifically designed to measure the signal on subject's forehead. Since I also has a Polar OH1 sensor, I thought it would be interesting to compare both a dedicated upper arm and forehead unit with the Hexoskin (looking at the raw signal for accuracy confirmation).
Here are the results.
This is a tracing of a very typical 3 minute interval done at about 360 watts, just above my VO2 peak power. The Polar is in red, Hexoskin in green with Power in blue. Although the correlation seems close toward the end, the beginning is certainly not.
Here is a Wingate 60 (max 1 minute power). This particular type of interval has always given an optical monitor the most difficulty, and I have yet to use one that could properly get the initial phase. Although there is a slight wrinkle in the Hexoskin at about 20 seconds, the raw data was reviewed and the signal was perfect (an advantage of having the raw data). The red is the Polar unit which basically misses the entire interval.
Why does it miss so badly?
Probably a combination of poor perfusion and motion:
Here is a tracing of Biceps O2 saturation (same intervals but different day):
The desaturation is much more pronounced and occurs faster in the Wingate 60 than the 3 minute fast start interval.
A closer look at this was done on a different day, but same interval type (first one is 350 x 3 min, second is a Wingate 60):
The sensor were placed in two locations, each being a recommended spot for the Polar OH1. The gray area is the medial forearm, green is the medial upper arm. The relative forearm area hypoxia (compared to upper arm) was deeper in the 3 min interval (near zero forearm vs 20% upper arm), but both sites approached zero during the Wingate 60. This may be one reason why the Polar unit was able to track during the 3 minute session, but not the all out Wingate 60.
The following is data from the Hexoskin motion data field:
The light blue line is motion and is much higher in the Wingate 60, perhaps by a factor of 2.
Therefore both tissue hypo-perfusion and severe motion are both contributing factors to the inaccuracy of arm based optical heart rate measurement.
The Moov Sweat:
This optical sensor is designed to be worn on the forehead with the supplied case and headband. According to the designers, it should provide a more accurate reading since motion is less than arm based locations as well as monitoring a site of preserved (facial) tissue perfusion even at high intensity. Aside from some technical issues in placing this under a cycling helmet, it is very similar in appearance to the Polar (but slightly thinner).
How does it perform?
Here is the 3 min 360 watt interval, with the Moov in red, Hexoskin in green. The results are quite close, with some minor deviation toward the end (of just a few BPM+-). Of importance though is the tight fit over the first minute, something not seen with the Polar.
This is amazingly close both early, middle and post interval recovery. I was extremely impressed and never thought that this was possible with an optical sensor. There must have been some motion given how hard I was swinging in the saddle as well as the helmet tapping the sensor with each sway.
The individual data points are presented below:
The correlation is impressive!
As another confirmation, here is a short 25 second maximal burst later that same session:
Still pretty spot on.
For me, there are still some kinks to work out. The main one is how best to place this under the helmet. If one was using this to run, swim or ski, there may not be any issue. For now I am placing this high on the forehead, above the helmet anchor band. The first tracing above (3 minute interval) used a placement below the anchor, just above the eyebrow. This was not very comfortable and I would not recommend it.
One of the Hexoskin's strengths is the ability to look at the raw EKG signal and make sure it is solid before starting your workout. Does the Moov Sweat have this ability? It just may.
In the smartphone app there is a screen showing the signal strength and wave form.
Here are some examples:
Center of forehead, relatively low amplitude waveform.
Lateral forehead low:
Better waveform than above
Left forehead, high near the hairline:
Very good waveform with high amplitude
In short, there is visual confirmation of signal quality and waveform. This can aid in placement and potentially (if the company enhances the software) validate peak heart rates if the raw data is recorded.
Summary and Conclusions:
- Measurement of heart rate continues to be an essential part of ongoing assessment of peak fitness, VO2 peak, zone training, recovery and over reaching.
- In order to measure this with the most precision, a device should have excellent correlation to an EKG signal.
- This correlation should be present especially during high/peak intensities that are associated with poor tissue perfusion and severe jerky motion.
- The motion and perfusion issues interfere in optical wrist and arm based heart rate monitors. The measured saturation in the recommended Polar unit locations approached zero during an all out maximum 60 second effort. In addition, the motion sensor data indicated almost twice the motion in the Wingate 60 than a 3 minute VO2 peak interval.
- A forehead based optical sensor seems to overcome this limitation. In my examples, the Moov device correlation was excellent compared to a 1 lead EKG (Hexoskin shirt). The improvement in precision (compared to arm based) was most evident at high intensity with severe body motion, the critical area of interest for physiologic measurements.
- Under perfect conditions the accuracy of belt type units also rivals that of the EKG.
- Unfortunately, if there is less than optimal electrical physical contact or introduction of other artifacts, there will be signal distortion causing derived heart rate inaccuracy.
- Although in many cases the blind faith in a belt sensor's precision does seem reasonable, there is no absolute guarantee that this is correct. Even the Hexoskin can record poor quality data, but we can assess this by looking at the raw tracing.
- If an athlete needs the highest quality heart rate data, only a device with raw signal recording capability should be considered.
- For high intensity exercise, the forehead based Moov Sweat will serve as an accurate heart rate monitor comparable to a belt unit.
Part 2 of the Moov HR review