Understanding the Quality of Muscle Oxygen Metrics

Posted by Roger Schmitz on Sat, Feb 13, 2016 @ 09:02 AM

Understanding_the_Quality_of_Muscle_Oxygen_Metrics.pngMoxy was the first muscle oxygen monitor developed specifically for athletes. Originally conceived as a possible solution for medical applications like Acute Compartment Syndrome, Peripheral Arterial Disease and Heart Failure, it has been on the market for over two years supporting trainers and athletes globally in their quest to improve sports performance. We’re starting to see competitive devices enter the market; a welcome sign of the market recognizing the usefulness of the technology for athletes. However, it does start to bring up questions about differences between these devices and the quality of their respective muscle oxygen metrics.

The purpose of this blog post is to provide information on which aspects of a muscle oxygen monitor’s measurement capability are critical to consider for all serious coaches and athletes who are trying to understand physiologic limiters and compensators through real-time, physiologic-based training.

In the process of designing and evaluating NIRS systems for measuring oxygenation for the past thirteen years, I’ve learned which factors are important for accuracy—many of which have proven difficult for other devices to overcome. For the past three years, we here at Moxy have been fortunate to have worked very closely with leading physiologists like Juerg Feldmann as well as top elite-level coaches and trainers to evaluate, modify and ultimately vet the Moxy Sensor. Our colleagues have helped us ensure it meets the specific needs required for physiologic-based training.

The seven aspects below are critical requirements for muscle oxygen monitor performance. Many devices have difficulties with these requirements or are completely unable to support them.

1. Isolate the Oxygenation Signal to the Muscle Layer

2. Accurately Indicate the Full Range of Oxygen Saturation on a 0% to 100% Scale

3. Indicate Changes in Total Hemoglobin

4. Accommodate a Wide Range of Fat Layer Thickness

5. Work on a Wide Range of Muscles on the Body

6. Provide Readings that are Stable with Temperature

7. Robustness to Skin Features

1. Isolate the Oxygenation Signal to the Muscle Layer

When we place the sensor over a muscle, the underlying tissue consists of skin, fat and muscle. The skin contains a significant amount of blood that we need to measure through. Blood flow to the skin is regulated by the body primarily as a temperature control mechanism rather than in response to the load on the muscle. This is one of the major concerns that Juerg ran into with the older NIRS devices he was using when we first met. While testing, he would sometimes see the reported oxygenation values increase when every other instrument he had to measure respiration, muscle activation and blood flow indicated the muscle oxygenation should be going down. The reason for this was that the NIRS device he was using measured the combination of the skin and muscle. In those situations, the skin blood flow was increasing to help cool the athlete, which obscured what was going on in the muscle. This is why it’s critically important to know that you are truly measuring SmO2 in the muscle.

2. Accurately Indicate the Full Range of Oxygen Saturation

When we are measuring oxygenation on athletes, it is useful to know if oxygenation is rising, stable or falling, but this is clearly not enough. It’s also crucial to know where the saturation is on a scale of 0% to 100%. We’ve had cases of athletes going all out on a bicycle, and his muscle oxygenation drives nearly to 0%. We’ve also had cases of a runner going all out on a treadmill, and her muscle oxygenation doesn’t go below 80%. In both cases, they went down from their baseline, but the runner had a very different physiologic limitation than the cyclist. The runner wasn’t using all of the oxygen that was present in the muscle, while the cyclist was.

Many NIRS devices measure oxygenation, but they are not accurate on the scale of 0% to 100%. Even if they report a number between 0 and 100, they aren’t accurate on the real scale. Kevin McCully’s research group developed a technique of applying a tourniquet to the test subject to attempt to “calibrate” out devices with this problem. While this was a functional solution, it’s not practical in a training setting. Accurate readings are necessary for determining performance limiters.

3. Indicate Changes in Total Hemoglobin

NIRS devices are typically capable of making two independent measurements. However, some devices only report the oxygenation value and do not report the total hemoglobin value. Both values are critical for using muscle oxygen data for physiologic-based training. The total hemoglobin value doesn’t measure blood flow directly, but changes in the total hemoglobin value follow patterns that give powerful insight into what’s going on with blood flow.

4. Accommodate a Wide Range of Fat Layer Thickness

The fat layer really complicates the ability of a NIRS device to measure oxygenation in the muscle. Very roughly, the sensor can penetrate about ½ of the distance between the furthest space emitter and detector opotodes. However, the accuracy of the readings will be affected by the fat layer unless the algorithm properly accounts for this layer.

When testing athletes, it is common to measure on multiple muscle sites, and even lean athletes have a wide range of fat layer thicknesses on various muscles.

5. Work on a Wide Range of Muscles

Some NIRS sensors are designed to measure only on one specific muscle. While this may provide some advantages to making a universal attachment feature, the restriction isn’t worth this small benefit.  In reality, it’s very common to measure multiple muscles even for a sport like cycling, where the work is mostly done with the legs.  Of course, it’s useful to measure on various working muscles in the quads and calves, but it’s also very helpful to measure the response in a non-involved muscle like a deltoid or even a respiratory muscle like the intercostal. When you look at other sports like running or rowing, or for applications like strength training, you need to be able to place the sensor on any large surface muscle.

6. Provide Readings that are Stable with Temperature

This is a factor that is largely driven by the physics of LEDs and laser diodes. These are very useful light sources for NIRS because they are very bright, efficient, emit a narrow bandwidth of light and are reasonably affordable. They have one major drawback in that they change color with temperature.

Some NIRs devices “get by” with ignoring this problem because the sensor is attached to the human body, which helps maintain a steady temperature. However, there can still be significant issues while the device warms up, or if the device is on an extremity exposed to weather, or if the device is used under water.

It’s not trivial to correct for the temperature errors, because the effect is sensitive to oxygenation values, fat layer thickness, melanin concentration and other factors. Proper temperature correction is important to ensure proper interpretation of the data.

7. Robustness to Skin Features

Hairs, moles and tattoos can be concerns for muscle oxygen readings. However, if the sensor doesn’t accurately indicate the full range of oxygen saturation anyway, this won’t be of much additional concern. In truth, these types of effects generally just shift or alter the scale of the values; they rarely would make decreasing oxygenation show up as increasing oxygenation.

Since NIRS is fundamentally a color measurement, it’s not possible to be completely insensitive to unknown colors introduced in unknown ways in the optical path of the sensor. If oxygen saturation accuracy is a concern, then tattoos should generally be avoided. 

Careful design of the measurement hardware allows the muscle oxygen sensor to be robust to features that simply block part of the optical path like hairs and moles.


Understanding_the_Quality_of_Muscle_Oxygen_Metrics_image_2.pngHow Moxy Handles These Factors

The short answers are:

  • Meticulous Attention to Detail
  • Advanced Algorithmic Mathematics

Moxy uses an algorithm that is 100% based on a technique called Monte Carlo modeling. Essentially, this method uses a mathematical mode to trace how light propagates through tissue, one photon at a time.  This allows us to include all kinds of potentially confounding factors like melanin, fat layer thickness and skin blood flow and design the algorithm to accommodate them. The model treats the tissue as layers of epidermis, dermis, fat and muscle and allows us to vary the optical properties of each layer independently. Modeling and accounting for all of these sources of variation is the first part of the meticulous attention to detail.

There are a couple of serious challenges to implementing this method. First, it is wickedly computationally intensive. Some of the modeling runs take more than four days on a very high-end computer processor. No one wants to wait four days for a SmO₂ reading (although we’ve never tried it with IBM’s Watson).

The second problem is that, this method only works in one direction, and it’s the wrong one. If you input the optical properties, it will tell you how much of each color of light makes it to the detector. However, we want the Moxy Sensor to work the other way. The Sensor knows how much of each color light reached the detectors, and we want to use that to determine optical properties in the form of SmO2 and THb. 

The solution to both of these problems is where the clever mathematics comes in. We do as many of the calculations as we can only one time and reuse those results for all Moxy Sensors. Then, to accommodate the exact properties of each individual Sensor, we do another set of calculations as part of the calibration in our lab (that takes 30 minutes to run on a computer for every Sensor). Finally, we do the last little bit of computation when we take each reading, which we can do in less than two seconds.

The advanced algorithmic mathematics leads us to a second part of the meticulous attention to detail.  The mathematical equations allow us to make judicious design choices and to see what parameters must be tightly controlled in order to make the design very robust. This allows us to be insensitive to the part of the optical signal being blocked by a hair or mole and to predict the precise response to temperature changes under the wide range of conditions that the Sensor is intended to operate under.

The mathematical models for the Moxy are publicly available in our patents. Here’s a link to the US patent. A patent does nothing to guarantee efficacy by itself, but the fundamental intent of a patent embodies the notion that we agree to show people what we’re doing in exchange for the right to stop people from copying what we’re doing. This seems to be working. I’ve had numerous discussions with NIRS experts about how Moxy works. This open discussion, coupled with the thousands of athletes tested over the last two and a half years, offers strong direct and anecdotal evidence that we got these critical factors right on Moxy.

I’ll end this section with two notes of caution:

1. The requirements described here are for using muscle oxygen information for physiologic-based training. The quality requirements for a device that uses NIRS to measure lactate threshold are much less stringent. It’s very possible for a device to perform exceptionally well at measuring lactate threshold and still perform poorly with respect to these Be careful to make this distinction.

2. These factors are difficult to test and establish positive results; our team continues to learn something new every day. Juerg Feldmann has an extensive testing lab more advanced than most research facilities. He’s made comparisons with non-invasive cardiac output, VO2, capnometry, EMG, power, heart rate and respiratory frequency, and he’s tested NIRS devices for many more years than Moxy has been around. He has a very intellectually critical and requires huge sample sizes of his own tests to draw his own conclusions. This methodology enabled him to uncover and evaluate the issues described here. The work discussed on the Moxy Developers Forum is immense, but it’s just a fraction of the work that has been done and discussed in emails, seminars and labs. Be careful with studies (even from prestigious universities) that test just a few dozen subjects or only one small aspect. This can be great information, but often reflects only a small part of the story.


There are some critical performance factors that are required for a muscle oxygen monitor to be useful for real-time, physiologic-based training. Moxy deals with these factors much more thoroughly than any other NIRS device.

We welcome your feedback.

order your moxy monitor today

Topics: About Moxy and About Us