We will review the remaining montages not previously discussed in the two previous posts, best practices, the strengths and weaknesses of popular montages, montage selection strategy, and optimal display settings. You will gain more from this post if you read the two previous installments.
Click on our narrator icon to listen to this post.
Best Practices from the American Clinical Neurophysiology Society Guideline 3 (2016)
The Committee reaffirms the statements pertaining to montages set forth previously in the Guidelines of the American Clinical Neurophysiology Society (ACNS) and that are paraphrased as follows:
(a) that no less than 16 channels of simultaneous recording be used, and that a larger number of channels be encouraged,
(b) that the full 21 electrode placements of the 10-20 system be used,
(c) that both bipolar and referential montages be used for clinical interpretation,
(d) that the electrode derivations of each channel be clearly identified at the beginning of each montage,
(e) that the pattern of electrode connections be made as simple as possible, and that montages should be easily comprehended,
(f) that the electrode pairs (bipolar) preferentially should run in straight (unbroken) lines and the interelectrode distances kept equal,
(g) that tracings from the more anterior electrodes be placed above those from the more posterior electrodes on the recording page, and
(h) that it is very desirable to have some of the montages comparable for all EEG laboratories.
2.2 The Committee recommends a “left above right” order of derivations, i.e., on the recording page, left-sided leads should be placed above right-sided leads for either alternating pairs of derivations or blocks of derivations. This recommendation coincides with the prevailing practice of most EEG laboratories, at least in North America and in many other areas.
Perspective
The entire EEG field is rife with semantic disagreements. We have made the point that all montages as well as all sensor comparisons are referential
Additional Montages
We briefly touched on the average reference montage. Additionally, several montages are also in common use. One is the linked ears montage. This is one of the montages sometimes referred to as “referential” montages to distinguish them from the sequential bipolar montages. The difference is that each scalp electrode is assigned the positive (+ or active) condition and a single common reference is used for the negative (- or reference) condition in the common mode rejection comparison. In most cases, these montages could be called common reference montages rather than simply referential, which would help differentiate them from other approaches since all montages are essentially referential.
The linked ears montage is one of these common reference montages because each scalp electrode is compared to the sum of the two ear or mastoid electrodes. Another common reference is the common vertex reference – generally using the Cz electrode as the reference for all other scalp electrodes. The other frequently used common reference is the common average reference, where all scalp electrodes are averaged. This montage uses this result for the reference for each individual scalp electrode.
The image below, adapted from Lopez et al. (2017), shows examples of these three montages. Unlike this graphic, in most cases, the midline electrodes are also included in these calculations.
Caption: three common referential montages include: (A) the Common Vertex Reference (Cz), (B) the Linked Ears Reference (LE), and (C) the Average Reference (AR).
In the linked ears/mastoid reference, there is generally a calculation involving adding and subtracting the signals from the ears before the combined signal is used as the reference. Older systems used a physical connection (called a “jumper”) between the two reference electrodes, often causing a current flow between the two electrodes and distorting the recording results. Often a resistor was added to this jumper to inhibit this effect. The digital processing of each signal independently and the resultant mathematical derivation are not susceptible to such distortion.
Another referencing system in common use is the Laplacian montage. This is a montage approach that was not available with older analog systems. It is sometimes called a local average montage since it uses a subset of electrodes surrounding the electrode of interest to create a local average value to which the center electrode can be compared. This is generally thought to enhance the ability to visualize locally occurring events in the EEG while suppressing effects that are common to the area. This may include the suppression of drug or medication effects, but this suppression is not the total elimination of these effects, and this claim must be viewed with caution. Gordon and Rzempoluck (2004) suggest that this approach can enhance the visualization of focal discharges and improve localization.
A Closer Look
We will examine the surface Laplacian (SL), linked ears reference, average reference, longitudinal bipolar, transverse bipolar, circular bipolar, common vertex (Cz) reference, average reference, and linked ears reference montages.
Surface Laplacian (SL) Montage
The SL is based on some complex concepts and calculations. Rather than a simple averaging of the voltage potentials of the electrodes immediately around the electrode of interest, it is rather an attempt to define the electrical field around that electrode. The result is a current source density (CSD) measure of current rather than voltage. In EEG, this measurement is in microamperes. The current is proportional to the potential differences between every two combinations of points or electrodes, one being the center or electrode of interest and the other being one of the perimeter electrodes.
The result of the Laplacian calculations provides an estimate of the electrical field surrounding the electrode of interest and represents the current flowing toward or away from a given electrode. This is a measurement of current flow perpendicular to the cortical surface measured as the rate of change of the potential field gradient around the recording site (Gordon & Rzempoluck, 2004). Because this calculation is based on the average of the surrounding electrodes, common influences are reduced, and focal activity is enhanced (Carvalhaesa & Acacio de Barros, 2014).
There are several methods for calculating the SL, but that discussion is beyond the scope of this post. Please see the cited papers and Nunez and Srinivasan (2006).
The Laplacian montage is not affected by the ear/mastoid references as they aren’t included in the calculation. It also visualizes local detail that is often difficult to see in other montages. The SL suppresses the general effects of global EEG activity and medications/drugs in favor of what is happening immediately below the electrode of interest.
Some drawbacks of the SL include the edge effect, where electrodes at the edges of the measuring field, such as Fp1 and Fp2, F7 and F8, O1 and O2, and so on, only have adjacent electrodes on three sides, and therefore the calculation is less accurate.
The SL also appears to add EEG content to electrodes in some cases. However, this is difficult to demonstrate because every EEG visualization depends on various factors, not the least of which is the referencing system. Finally, EEG activity distributed over the entire scalp would not be seen in this montage; therefore, it is best suited to identifying local activity (Gordon & Rzempoluck, 2004).
However, the Laplacian montage does an excellent job of identifying localized activity and reducing the effects of common influences.
Here is an example of a weighted average Hjorth Laplacian analysis of the electrode at Fz. The electrodes immediately around in radiating circles are weighted by distance in terms of their contribution to the reference.
Caption: Each sensor is referenced to an average of all other sensors, "weighted" by distance from the sensor.
A more typical Laplacian montage with just the electrodes immediately surrounding the target electrode being used in the reference.
Caption: Each sensor is referenced to the average of the surrounding sensors.
Below are five examples of the same eyes open data viewed in Laplacian, linked ears montage, average reference, longitudinal bipolar, and vertex (Cz) reference. All images show mu rhythm at C3 and C4, but the Laplacian and average reference views show the mu rhythm more clearly. The Laplacian appears to differentiate the mu rhythm from the background most effectively. Note that the Laplacian montage uses a y-scale setting of 500 μA, while all the others use a 50 μV y-scale. Laplacian Montage
Linked Ears Montage
Average Reference Montage
Longitudinal Bipolar Montage
Vertex (Cz) Reference Montage
Linked Ears Reference Montage
The linked ears reference montage compares each scalp electrode to the combined signal from both ears or mastoid locations. This is in search of a "neutral" reference and some believe that this is the case. The benefit of the linked ears is better visualization of central/midline sources as well as frontal and prefrontal activity. However, linked ears are well known for adding cardiac activity, EMG activity from neck and jaw muscles, as well as EEG patterns such as alpha, theta, or transient activity to scalp electrodes. The example below shows this clearly in the circled epoch.
Average Reference Montage
The average reference montage uses an average of all scalp electrodes as the reference. This can help eliminate common sources and generalized EEG patterns in favor of activity that is more specific to each scalp electrode site. At the same time, it can minimize important EEG activity that exists at multiple sites, such as that in the example EEG. It can also contribute scalp EEG patterns to locations where they don’t actually occur.
Longitudinal Bipolar Montage
The longitudinal bipolar (longBP) montage follows the longitudinal sequence of electrode comparisons or derivations. The left-side electrode derivations of the more lateral (lateral frontal, temporal and parietal) locations precede the more medial (parasagittal) derivations, followed either by the center/midline if used or by the right side medial and then lateral derivations. Sometimes the central sequence is at the bottom. See the sequence of numbered derivations shown above. Good for comparing and visualizing differences in the left hemisphere compared to right hemisphere characteristics. See the example below, from the NeuroGuide database representing a sample provided within the software of an individual with a right hemisphere parietal impact injury. Note the marked differences between left and right hemisphere activity. This is a 10-second window and a 50 μV y-scale.
Transverse Bipolar Montage
The transverse bipolar montage follows the suggested anterior-to-posterior orientation, with prefrontal and frontal sequences of electrodes displayed first and the rest following.
Below is an EEG example from NeuroGuide showing the effects of a right parietal injury. This difference is most clearly seen in derivations involving P4 compared to those involving P3. This montage helps identify relationships that may not be clear in the LongBP montage.
Circular Bipolar Montage
The circular bipolar (CircBP) montage shows electrode pairs following a circular (coronal) orientation, often beginning with Fp1-Fp2, Fp2-F8 or beginning with T3-F7, F7-Fp1 and so on, following the left over right recommendation. This montage also can highlight activity and relationships that escape the other two montages.
The EEG tracing below shows the example EEG in the CircBP montage.
Common Vertex (Cz) Reference Montage
The common vertex (Cz) reference montage, generally using Cz as the common reference for all other electrodes, simply compares each scalp electrode to the same reference. It provides a common voltage in the reference channel. Benefits from this montage include being able to visualize electrodes that are equidistant from the vertex, such as Fp1 and Fp2, F3 and F4, O1 and O2, and so on.
Also, it allows us to compare their characteristics to reveal differences between hemispheres and between frontal and posterior areas. This montage can be useful when identifying interhemispheric amplitude asymmetries and other metrics. In the example EEG below, right and left differences can be seen. However, activity in the general vicinity of Cz can be added to other electrodes in some cases.
Average Reference Montage
The average reference montage uses an average of all scalp electrodes as the reference. This montage is also useful for identifying local activity, particularly in temporal lobe areas where the ear references may either contribute to or cancel the same activity. If the software can exclude electrodes affected by large EEG sources, then the resulting average excluding these sources will be more neutral. Graphic © learningeeg.com.
It is noted that the average reference skews phase and coherence calculations and is therefore not used for these particular database comparisons where the database was collected using linked ears as the reference. Z-scores can be calculated when the database provides norms specific to the average reference. The average reference montage is often cited as the "best" for viewing the EEG, but, like all referencing systems, the average reference has its own problems. Because it represents an average of the voltage at all electrodes, it contains, within that negative (reference) channel, a fixed voltage that can then affect the resulting EEG tracing on the screen (Nunez & Srinivasan, 2016).
Suppose this contribution contains high amplitude artifacts, persistent EEG patterns such as alpha, theta, delta activity, or any other factor. In that case, these can be added to those electrodes that don’t already have these patterns, which can cause the rejection of these patterns in electrodes that share them. The problem is that the effect is often quite diffuse and, therefore, more difficult to spot.
In the authors' practice, the average reference statistical (z-score) topographic displays (e.g., maps) often correspond quite closely to those derived from the Laplacian montage, whereas those from the linked ears/mastoid montage do not.
Linked Ears Reference Montage
The linked ears reference montage is one of the most commonly used references in neurofeedback. The original NxLink database developed by E. Roy John and the NeuroGuide database developed by Robert Thatcher collected ear electrodes when conducting EEG recordings to make linked ears montage computations. Calculations of phase and subsequent coherence measures derived from the phase calculations in these databases are made using the linked ears montage.
The linked ears montage is generated by comparing each scalp electrode to the average of the two ear/mastoid sensors. The choice of ears or mastoids does not seem significant, but there may be noticeable differences in individual cases. If the equipment used can access/record either or both the ears and the mastoids, the clinician can view the recording as it is occurring while using one and then the other and then choose the best one for that recording session. The presence of pronounced electrical activity in the reference channel and/or individually in each reference sensor would suggest against using that source in favor of the choice with the most minimal contribution to the resulting recording.
The ear reference montage has sometimes been conducted using only one ear or mastoid, and arguments have been made supporting the ipsilateral (same side) versus the contralateral (opposite side) reference. This can be a consideration when using a small number of scalp sensors, but when using all 19 of the 10-20 scalp electrode sites, it is more common to use the two ear/mastoid sensors in a linked approach. Some guidelines suggest that the angle of the mandible may also be used (Acharya et al., 2016).
The linked ears/mastoid is a good montage for viewing EEG activity in the central electrode locations, along the vertex particularly, because they are quite far away from the reference. Therefore there is only minimal common activity subject to rejection. However, it is not as helpful for temporal electrode locations because these often share EEG activity with the references, which is then rejected. Again, the linked ears reference often contributes alpha activity to frontal and central signals due to the retention of anything different in the common mode rejection process. Note the widespread alpha activity seen in the linked ears montage in the images above, showing mu rhythm examples that do not appear in the other montages.
The linked reference montage does simplify the visualization of the EEG tracings (Valentine, 2020). Since there are no bipolar comparisons, there are no phase reversals in the sense of the bipolar montage. However, the field of activity around a large EEG event, such as an epileptiform discharge, can sometimes be identified in this montage. All the negative EEG activity at each electrode site shows a negative (typically up) deflection in the EEG tracing, and positive electrical shifts deflect down. This is, of course, in reference to the linked ears/mastoid or another common reference.
As mentioned, there are other choices for references, including the vertex reference, usually at Cz. The choice of reference is important because it affects what can be seen in the resulting EEG tracing. The vertex Cz location is a good choice if temporal lobe epilepsy is an issue. Using linked ears/mastoids for viewing this activity could result in important data being rejected as the same patterns may show up in the reference as in temporal lobe sensors. Another issue with ear/mastoid electrodes is that they often contain artifacts such as EMG from the jaw or neck muscles as well as ECG (cardiac) artifacts that can then be contributed to all scalp sensors. The problem of reference contamination with alpha activity from the temporal lobes has already been addressed.
However, a Cz reference would not be a good choice if the client is sleepy because sleep indicators such as vertex sharp waves may be attenuated. Finally, the ear/mastoid references would be better than Cz if the goal is to visualize frontal EEG activity, such as frontal slowing or the effect of a frontal impact injury. Cz would also not be a good choice if there is distinctive EEG activity at or around this site since that activity could be added to other sensor locations. In the vertex reference image above, it appears to add this pattern to other sensors because there is likely mu rhythm at Cz due to its close proximity to C3 and C4.
Selecting a Montage
The choice of montage is, therefore, not a simple one, and often the best advice is to use at least four different montages, adding more as needed.
The minimum approach is, to begin with, either the average reference montage or one of the bipolar montages, usually the longitudinal bipolar montage (LongBP). This is followed by the Laplacian montage and finally the linked ears/mastoid montage.
One benefit of beginning with the LongBP montage is that it is commonly used in EEG textbooks and EEG atlases. When comparing a client’s EEG features, particularly when looking up a particularly unusual pattern or something you are unfamiliar with, it is essential to use the same montage used in the reference source, such as Stern’s An Atlas of EEG Patterns (2004). Spike and wave patterns, inter-ictal patterns, generalized slowing, and all the wonderful patterns that have been reduced to acronyms such as POSTS, BETS or BSS, SREDA, FIRDA, PLEDS, and BIPLEDS, and more. Because many of these patterns have visual similarities to epileptiform activity, it is common for beginners to mistake them for actual indicators of epilepsy, which is why purchasing and regularly studying an EEG atlas is so important.
If starting with the bipolar montage, it can be useful to compare the LongBP montage with either the transverse bipolar (TransBP) or circular bipolar montage (CircBP) because they can provide views of other derivations for electrodes of interest, which can help with localization. Transverse Bipolar Montage
Circular Bipolar Montage
Of course, identification of epileptiform activity can only be made by a qualified neurologist. Still, the neurofeedback practitioner can identify what appears to be abnormal activity and make an appropriate referral.
Following the bipolar montage, the average reference montage can be the best choice for identifying artifacts using a normative database that relies on the manual selection of artifact-free data.
Independent Component Analysis (ICA)
Database software that provides automatic artifact removal methods like independent component analysis (ICA) will still require the clinician to view the EEG recording using multiple montages to identify features of interest.
While ICA is controversial in the neurofeedback community, it has been validated in various studies and produces clean, artifact-free results. It was found to have minimal effect on the resulting analysis (Iriarte et al., 2003; Urrestarazu et al., 2004). However, Thatcher states in an undated paper that ICA artifact removal results in distortion of phase information in the EEG and therefore results in an invalid analysis of coherence derived from phase and subsequent network analysis. This is supported by Nunez and Srinivasan (2006), and though amplitude/power and relative power, ratios, and peak frequencies can be reliably calculated, phase and coherence cannot when using ICA.
One caution noted by Dorfer (2020) is that ICA works well with standard eye blink and eye movement artifacts but is less effective for random movement artifacts and other transient artifacts, such as electrode pop.
Montage Selection Strategy
Obviously, our goal when viewing the EEG is to identify areas that deviate from typical behavior, correlate those differences with client symptoms, and design a training protocol or group of protocols to address those differences.
When we identify an EEG feature representing slowed alpha, high amplitude frontal theta activity, excess fast activity, lack of a typical alpha response, or any other finding in one montage, we want to verify and validate that finding using additional montages.
Following the average reference montage, the Laplacian montage can be useful for further zeroing in on the areas of interest.
Finally, the linked ears montage must be consulted if only to identify areas of likely reference contamination that may affect subsequent topographic z-score maps, phase, coherence and network analyses, and other downstream evaluations.
Use Consistent Settings
One important consideration for viewing the EEG is to use consistent settings. The EEG is often viewed using a 50 μV y-scale (vertical axis) and a 30 mm per second x-axis (horizontal) equivalent ‘chart speed’ display.
In the early days of EEG, tracings were drawn by pens suspended over moving chart paper. A typical speed for adult EEG was 30 mm per second and for pediatric EEG was 15 mm per second. Now that almost all EEG is digital, similar equivalent displays show the EEG in this format. This is so the waves appear consistently the same every time they are viewed and can also be compared to reference sources. If different visual display settings are used, then it is difficult to identify wave patterns and abnormally high or low voltage values by visual inspection.
Most modern EEG software uses display time settings indicated in seconds rather than chart speed equivalents. In the NeuroGuide database, the setting that appears most similar to 30 mm per second is 10 seconds per page. Also, in this program, the y-scale setting adjusts automatically depending on the highest voltage present in the recording. It doesn’t allow for a constant setting, though the desired setting can be made manually each time the montage is changed.
Here are some examples of the same page of EEG activity viewed in the LongBP montage using various time and voltage scale settings. You will see that some views allow for good data resolution, and some do not.
This is the setting that NeuroGuide defaulted to with the display time at 6 seconds and the voltage scale at 30 μv.
Here the display is set to 3 seconds with the same vertical setting of 30 μV.
This example shows the same data with the time scale set at 10 seconds using the same vertical scale.
This shows the 10-second time window along with a 50 μV vertical scale. This fairly standard view can be compared to an EEG atlas.
This shows the 10-second time scale with the y-scale set to 200 μV, obviously losing waveform definition.
The same 10-second window with the voltage set at 1 μV does not allow proper visualization of the EEG.
The same data displayed using a Laplacian montage. Note that the y-scale is in μA (microamperes) rather than microvolts. A 300-μA scale is a typical resolution in the Laplacian montage.
These examples help to illustrate how important using the same scale settings can be. Whenever possible, use consistent scale settings and become familiar with them so that you can train your pattern recognition system to identify typical and unusual EEG features at a glance. This takes time and practice but is greatly facilitated by using the same settings.
When working with an EEG with very high voltages of EEG, that are not related to an artifact such as eye blink (see the example of eye blink artifact below), it may be necessary to use a higher voltage setting. However, it is important to use some consistent settings, such as 75 μV or 100 μV, so that you can also train yourself to see EEG features at those settings.
Eye Blink Artifact
Below is an example of a high-amplitude EEG of a young child requiring a y-axis setting of 75 μV for good resolution. 100 μV would also be acceptable.
Summary
This installment of our montage blog shows additional montage examples and discusses the uses, benefits, and limitations of each montage. The chart below attempts to organize that information into a usable tool for the clinician.
Glossary
amplitude: the strength of the EEG signal measured in microvolts or picowatts.
artifact: false signals like 50/60Hz noise produced by line current.
average reference montage: EEG recording configuration using an average of all scalp electrodes as the reference.
channel: the collection of three electrodes, the electronics that compare them, and the resulting output.
circular montage (CircBP): EEG recording configuration in which electrode pairs follow a circular (coronal) orientation, often beginning with Fp1-Fp2, Fp2-F8 or beginning with T3-F7, F7-Fp1, and so on, following the left over right recommendation.
common vertex (Cz) reference montage: EEG recording configuration that generally uses Cz as the common reference for all other electrodes, comparing each scalp electrode to the same reference.
common-mode rejection ratio (CMRR): the degree by which a differential amplifier boosts signal (differential gain) and artifact (common-mode gain).
derivation: assigning two electrodes to an amplifier's inputs 1 and 2. Montages combine derivations.
differential amplifier (balanced amplifier): a device that boosts the difference between two inputs: the active (input 1) and reference (input 2).
ear ground/reference: used for one or two scalp sensors. When only one or two "active" scalp electrodes are used, ground and reference electrodes are often situated on the earlobes.
earth ground: an electrical and physical connection to the earth.
edge effect: a drawback of the surface Laplacian montage in which electrodes at the edges of the measuring field, such as Fp1 and Fp2, F7 and F8, O1 and O2, and so on, only have adjacent electrodes on three sides, and therefore the calculation is less accurate.
EEG artifacts: noncerebral electrical activity in an EEG recording can be divided into physiological and exogenous artifacts.
electrocardiogram (ECG) artifact: contamination of the EEG signal by cardiac electrical activity.
frequency (Hz): the number of complete cycles that an AC signal completes in a second, usually expressed in hertz.
gain: an amplifier's ability to increase the magnitude of an input signal to create a higher output voltage; the ratio of output/input voltages.
ground electrode: a sensor placed on an earlobe, mastoid bone, or scalp that is grounded to the amplifier.
ground/system reference: an electrode that provides a return pathway back to the amplifier.
independent component analysis (ICA): a complex mathematical procedure for identifying and removing the artifacts contaminating an EEG signal.
International 10-10 system: a modified combinatorial system for electrode placement that expands the 10-20 system to 75 electrode sites to increase EEG spatial resolution and improve the localization of electrical potentials.
International 10-20 system: a standardized procedure for placing 21 recording and 1 ground electrode on adults on adults to provide a total of 19 channels. This system is used for typical 19-channel qEEG recordings, using 19 "active" electrodes, "reference" electrodes at A1 and A2, and a ground electrode.
Laplacian montage: EEG recording configuration in which a subset of electrodes surrounding the electrode of interest creates a lo