The do's and don'ts of baselines

A common question we get from our customers is what is a baseline and how to use it. Generally with fNIRS, the absolute values are arbitrary. You are always interested in changes relative to another point in time, for example if you investigate a change in oxygenation in response to a stimulus, you are actually investigating the change in response between a short moment before the stimulus was presented and the period in which it was presented. The period before a stimulus is often referred to as the baseline. In this blog we will describe the do's and don'ts of baselines.

What is baseline and how to use it?

For those just getting started with using NIRS it can be somewhat unclear what measurements mean. To explain this lets go back to the origin of the baseline.

Most commonly used NIRS is ‘continuous wave’ (CW) NIRS, which is based on the modified Lambert Beer Law. It uses a continuous firing light source. The light will enter the tissue and is both scattered (changing its direction) and possibly absorbed.  Both scattering and absorption are a cause why CW-NIRS provides relative measurements. In the figure below the normal Lambert-Beer Law is shown, wherein absorption in a cuvette is displayed. In the modified Lambert-Beer Law an extra term is added to compensate for scattering in the tissue.

 Absorption in a cuvette as described by the Lambert-Beer Law

Absorption in a cuvette as described by the Lambert-Beer Law

The scattering effect is clearly visible when you put your finger in front of a red laser pointer. You will see your whole finger light up, not just a straight line; the light is scattered in the tissue. So, if you would measure directly opposite of the laser on your finger, you would not receive 100% of the emitted light by the laser pointer, even if there was not absorption at all. 
Luckily, we can make the assumption that this scattering in all directions is constant, so if we measure a change in received light, there must a change in absorption in the path that the light followed from the source to the receiver.

Most absorption of NIR-light is by haemoglobin, the very reason why we can utilise NIRS for our measurements. However, the light is also absorbed by other chromophores (e.g. adipose tissue, hair, skin). Again, we make the assumption that this absorption is constant during the measurement and any change in received light is caused by a change in absorption by haemoglobin.

So, CW-NIRS uses a change in received light to calculate a change in concentration of haemoglobin. Therefore the change is always relative to a "starting situation" or baseline.

What does that mean?

CW-NIRS can only measure a change in concentration, therefore it cannot provide you with the starting concentration (so commonly set to an arbitrary zero). However it can provide you with a change from this baseline caused by any kind of intervention (for example contracting a muscle or increased activity in a certain part of the brain). This change is quantified in micromolar (which is micromole haemoglobine per liter of tissue). This can be a negative change, which means a decrease in concentration since the intervention.

How should I use a baseline?

A good baseline can be more difficult than you think. You can use either one time point in your data and set it to zero (by subtracting that value from all data points of that trace), or you can take a baseline for a certain time (e.g. 1 minute) and subtract the average of that from all data points of that trace. If the protocol allows you, it can also be good to record another baseline at the end of your measurement. If the baseline is not significantly different from your initial baseline this gives extra confidence in your data quality, or you might be able to use it for artefact correction.

NEW FEATURE IN SOFTWARE: use period for biasing?

The baseline should be similar to the intervention you want to test, only without that intervention. This seems trivial but experience learns that this can be a challenge. For example during muscle measurements during cycling, a certain resistance is compared to baseline where the subject was sitting still on the bike. This is feasible if you at least make sure that you measure the muscle with the leg in the same position for each subject. However, you might understand that if the position of the leg (pedal down or up) can make a difference for your baseline. It is OK because if you use the same position for all the participants, you have the same ‘starting position’ each time. You do need to be aware of this if you compare with literature. A better solution would be to use cycling at zero Watt for a minute or so and use this as baseline. Now the only difference is the increase in power of the cycling.

Another example for functional NIRS is what you have the subject do when you are recording the baseline. For example when you use certain memory or calculation tasks you must be sure that the subject actually stops his task during the rest/baseline periods. This can be achieved by offering a point to focus on or giving them another task of which you are certain it will not lead to any extra activity in the region that you are focusing on.

How can I use relative concentrations?

Functional NIRS (which part of the brain gets more active whenever a certain stimuli is given) commonly uses multiple channels. Basically, this enables you to see if just a few channels (hopefully the ones over the region of interest) are significantly different than the others. This can then be compared between subject groups.

In some (quite rare) cases researchers just do a resting measurement. They do not use the average values (as these are arbitrary), but they might use the fluctuations in the signal and perform (frequency) analysis. This can be within one signal, or they compare frequencies in multiple channels.

For muscle measurements performing an arterial occlusion might be helpful. Sometimes this is used as a kind of calibration. If the concentration changes reach a certain plateau this is assumed to be a minimal oxygenation state. The maximum shortly after releasing the cuff is assumed to be a maximum oxygenation state. Any changes caused by your actual intervention can be given relative to this minimum and maximum.

Next to this calibration, (venous and/or arterial) occlusions can be used to quantify oxygen consumption, blood volume, blood flow and reoxygenation rates. For more information you are referred to the thesis of Mireille van Beekvelt.

Others do a ‘baseline intervention’ (e.g. a functional task or a sprint) and then do the actual intervention (take a pill, do meditation, follow a certain training) and then do the same intervention as the baseline. Is the change by the second intervention different from the baseline intervention? Do be aware that in this case it would be good to use a placebo as well.

The don’ts of baseline

Do not compare averages of baselines between groups. The average is arbitrary and most commonly set to zero. Do not have too much artefacts (e.g. removing the sensor) between baseline and intervention. Therefore in fNIRS there is commonly a resting period before each stimuli which can be used as baseline.

Lets sum it all up

For easy reference a small list of do's and don'ts for baselines is compiled below:

Do's:

  • Have a baseline period before your intervention, preferably also after the intervention
  • Be careful with artefacts between the baseline period and the intervention
  • The baseline situation should be as identical as possible to the intervention situation

Don'ts

  • Do not compare the averages of baselines between groups
  • Do not remove the sensor between baseline and intervention