Lab 8: EEG 1

Introduction:


An electroencephalogram (EEG) is a currently used biomedical recording of a
summation of the electrical signals produced by many neurons that are firing over a period
of time. It is measured from the surface of the scalp. More specifically, the EEG signal is the
summation of action potentials and postsynaptic potentials of individual neurons in the
cerebrum of the brain. The EEG relates specifically to brain function because these
biopotentials are created from potential differences in the neurons within the brain as part of
its emotional, mental, and physiological function. The human brain consists of millions of
neurons that communicate information to one another and to the rest of the body utilizing
electrochemical potentials. Communication between the brain and the rest of the body is
made through both ascending and descending tracts of nerve fibers in the spinal cord.
These tracts contain two types of nerves: afferent and efferent. It is through these nerves
that the brain and body exchange information. The ascending tracts are made of afferent
nerve fibers, which are used for sending sensory information from various parts of the body
to the brain. In contrast, the descending tracts are made of efferent fibers, which send motor
movement messages from the brain to the target location in the body; these messages are
the mechanism in which the body responds to environmental stimuli.
Action potentials and postsynaptic potentials are the two mechanisms used in the
brain responsible for creating the EEG signal. The two differ based on their order of
occurrence. The action potential is the signal propagation through the neuron. In contrast,
the postsynaptic potential is the change in transmembrane potential following the release of
neurotransmitters at the end of a presynaptic axon. Additionally, the two potentials differ in
their mechanisms. While action potentials are all-or-nothing signals involving sodium and
potassium ions, the postsynaptic potentials are graded signals that control the amount of
neurotransmitter released where the axon of the presynaptic neuron meets the next
neuron.
The EEG signal is composed of four major component waves. These are the alpha,
beta, delta, and theta signals. The four signals are separated by their typical frequencies
and have some levels of amplitude overlap due to synchronous or desynchronous
potentials. Since the EEG is a summation of these potentials, a deceptively small amplitude
can be seen when potentials are asynchronous since the amplitudes are leading to some
cancellation. If the potentials are synchronous, then the EMG will have a higher
amplitude. Alpha waves (8-13 Hz, 2-100 uV) are the most common and have been found to
derive from the occipital lobe and tend to only be present when the subject is awake and in
a quiet state with closed eyes. They can also be detected from the frontal and parietal lobes
to a lesser degree. Beta waves are recorded from the parietal and frontal lobes. They
typically appear when the subject is mentally engaged or is attentive to an external stimulus.
Beta waves (13-22 Hz, 5-10uV) differ from alpha waves in that they are more
desynchronous and therefore sum to lower amplitudes on the EEG, called alpha
block. Theta waves (4-8 Hz, 2-10uV) occur during times of emotional stress and originate
primarily from the parietal and temporal regions of the brain. They are also associated with
degenerative brain diseases. Delta waves (05.-4 Hz, 20-100uV) occur only during deep
sleep or when the subject has certain brain diseases. These waves are unique in that they
are not associated with one particular lobe but from pyramidal neurons across the cerebral
cortex. However for subjects with illnesses, delta waves have been shown to originate in the
left temporal lobe of patients who suffer from diseases like temporal lobe epilepsy.
The EEG electrode placement is performed using the 10-20 system. This refers to
placing EEG electrodes on the scalp of the subject at approximately 10-20 percent
distances from one another proportional to the subjects head size. The electrodes are
spread out using this technique and cover all four major lobes. The electrodes are
designated with both a letter and number. The letters display whether the electrode is
taking readings from a lobe or from the central area of the head. The O letter designation,
for example, corresponds to the occipital lobe where the alpha waves can be
detected. Similarly, the F letter corresponds to the frontal lobe, P to the parietal lobe,
FP to both the frontal and parietal lobes, T to the temporal lobe, and C for the central
location of the head. In the experimental set-up, only the O and FP electrodes were
placed, in addition to electrodes on the mastoid processes labeled A. An alpha numeric
system is used to designate whether the electrode is on the subjects left side (odd
numbered) or the subjects right side (even numbered).
There are several noise sources that can interfere with the measurement of the EEG
signal. Among these is the typical 60 Hz noise that is found in most biopotential readings
from surrounding electrical systems, as well as other signals from nearby muscles and body
parts. The muscles that control the eyes and neck can have serious signal contamination on
the EEG. EOG (electrooculographic) from the eye and EMG (electromyographic) signals
from the neck and face, and even large body movements can create noise on the EEG. An
example of this can be seen in both blinking and chewing. These activities produce
undersired noise in the EEG and must be filtered out. As explored in earlier labs, this noise
is expected due to the fact that the human body utilizes electrochemical potentials within the
same ranges to create movement and maintain regular functions.


Experimental Results and Analysis:


Seven gold cup electrodes were placed on Subject B09. Four of these corresponded to
Fp1, Fp2, O1, and O2 on a standard 10-20 system of electrode placement. Two were
additionally placed on the mastoid processes, and corresponded to A1 and A2. The final
electrode was placed on the center of the forehead as the ground. The raw EEG signals
from Fp1, Fp2, O1, and O2 were plotted over time in Figure 1 below. The typical
appearance of a raw EEG wave has a more evident peaks and this characteristic shape is
absent from the figure below. From these four signals, it can be seen that a noise is
present in the recordings, likely at a higher frequency and amplitude in order to drown out
the EEG wave. This noise was accounted for in subsequent filtering.


Figure 1. Raw EEG Signals for Two Frontal (Fr) and Two Occipital (Oc) Channels over
Time


In addition to the noise that was present from Subject B09, the noise associated with
blinking and chewing can interfere with the EEG signal. These are electrooculographic
(EOG) and electromyographic (EMG) noise, respectively. To characterize this EOG noise,
the subject was asked to blink frequently. The recorded data can be seen in Figure 2. The
high frequency noise is once again present, leading to an indistinct EEG wave. The blink
noise can be seen in the peaks of the graph, when compared to the raw EEG without
blinking from Figure 1.


Figure 2. Raw EEG Signal for Channel Fr1 with EOG Noise


The EMG noise was characterized similarly, with the subject chewing as the EEG signal
was recorded. The effect of this noise on the EEG signal can be seen in Figure 3
below. Again, the high frequency noise obscures the characteristic EEG signal, and the
EMG noise can be seen in where the plot peaks.


Figure 3. Raw EEG Signal for Channel with EMG Noise


In order to examine the alpha waves of Subject B09, the reading from Channel O1 was
observed and filtered. Since the alpha waves are associated most strongly with the
occipital lobe, Channel O1 and Channel O2 were considered for this portion of the
laboratory. However, Channel O1s EEG signal appeared to contain less noise and was
thus selected for further filtering. A bandpass filter was applied, and the results can be seen
for the time domain in Figure 4 below. In terms of amplitude and distinct peaks, the alpha
waves were much more distinct when Subject B09s eyes were closed.


Figure 4. Time Domain EEG Signal for Channel O1 with Bandpass Filter for Open Eyes
(top) and Closed Eyes (bottom)


The filtered data was also viewed in the frequency domain, as can be seen in Figure 5
below. For the EEG signal with open eyes, the estimated peak frequency is 6.6906 Hz, a
frequency more characteristic of a theta wave. With closed eyes, the EEG signals
estimated peak frequency is 9.2116 Hz, which is much more characteristic of an alpha
wave. The DC noise and 60 Hz equipment noise is visible in both of the EEG frequency
plots.


Figure 5. Frequency Domain EEG Signal for Channel O1 with Bandpass Filter for Open
Eyes (top) and Closed Eyes (bottom)


Since alpha waves are not the only waves of interest, a data file with Subject B09s eyes
open was processed for beta wave activity. The data file with the eyes open was selected
because the beta waves are associated with an awake, engaged subject. The data was
processed with a bandpass filter for the frequency range characteristic of beta activity in
Figure 6 below. Channel Fp2 was selected for processing because the beta waves
originate from parietal and frontal lobes of the brain; Channel Fp1 had more noise in
comparison. It can be seen from the frequency domain plot that the signal was restricted to
frequencies within the beta wave range; thus, if any noise remains in the signal it would
have to have a similar frequency to a beta wave.


Figure 6. Filtered EEG Signal for Channel Fp2 for Time Domain (top) and Frequency
Domain (bottom) for Beta Wave Activity


A similar analysis was performed for a data file where Subject B09s eyes were
closed. Alpha waves were the activity of interest for this data file, and Channel O2 was
selected over Channel O1 due to a lesser amount of noise. This can be seen in Figure 7
below. As evidenced by the frequency domain plot, the DC noise and 60 Hz equipment
noise were successfully removed from the signal. In addition, the peak frequency was
within the expected range for alpha waves and any remaining noise had a similar frequency
to the alpha waves.


Figure 7. Filtered EEG Signal for Channel for Time Domain (top) and Frequency Domain
(bottom) for Alpha Wave Activity


Another key component of processing data is the removal of noise. For the data recording
involving the EOG noise, the time and frequency domain graphs of the raw EEG signal can
be seen below in Figure 8. In the time domain plot, the signal of interest is obscured by the
noise, which can be examined from the frequency domain. DC and 60 Hz equipment noise
is present, in addition to a noise of unknown frequency that has caused indistinct signals
prior to this. Additionally, since the peak at 1.8 Hz is higher than the surrounding peaks, this
frequency corresponded to the EOG noise.


Figure 8. Raw EEG Signal with EOG Noise in the Time Domain (top) and Frequency
Domain (bottom)


As seen in Figure 9, a bandpass filter was applied to the raw signal in order to remove the
DC noise, EOG noise, and 60 Hz noise. After applying the filter, the signal was more
distinct and was closer to the characteristic appearance of an EEG wave. The frequency
domain graph confirms that the noise present in the raw signal has been removed.


Figure 9. Filtered EEG Signal with EOG Noise in the Time Domain (top) and Frequency
Domain (bottom)


As with the raw EEG signal with EOG noise, the EEG signal for the EMG noise was plotted
in frequency and time domain in Figure 10. In the time domain, it can be seen that noise is
obscuring the characteristic EEG signal once again. By referring to the frequency domain
plot, it can be seen that the DC noise and 60 Hz equipment noise is present. In addition,
the peak at 1.8 Hz is likely the EOG noise due to blinking once again and also needs to be
filtered. The peak at 4.2 Hz is the EMG noise due to chewing, since the peak is significantly
higher than that of the surrounding frequencies. The peak at 2.1 Hz may be a combination
of the two previously discussed noises, or may be motion noise due to head movement.


Figure 10. Raw EEG Signal with EMG Noise in the Time Domain (top) and Frequency
Domain (bottom)


As seen in Figure 11, a bandpass filter was applied to the EEG signal in order to remove
the noise discussed previously. After performing this filtering, the time domain signal
appeared to have a more characteristic EEG signal. The frequency domain plot confirms
the removal of the DC noise, 1.8 Hz noise, 2.0 Hz noise, and 60 Hz noise. The 4.2 Hz noise
may not have been completely removed, but its relative amplitude with respect to the rest of
the EEG frequencies is much less significant, allowing for a clear reading of the EEG signal.


Figure 11. Filtered EEG Signal with EMG Noise in the Time Domain (top) and Frequency
Domain (bottom)


Discussion:


There were several sources of noise seen in this lab exercise. As expected, 60 Hz
noise from nearby electronics, including the BioRadio itself were observed. Other sources of
noise included the subjects blinking and chewing movement. As explored in Figures 2 and
3, subject B09 was asked to blink and later to chew in order to reference exactly how the
signal was affected by each. These sources of noise could be seen as sharp peaks on the
time domain as well as small and large peaks across the frequency domain. Another source
of noise in the signal included a frequency present above 25 Hz. Its source was of unknown
origin and had to be filtered out with the bandpass filter in order to process the desired EEG
signal. It was best observed in Figure 1, where it obscured the characteristic EEG signal. As
seen in Figure 10, the signal displays the raw signals with EMG noise artifacts that had to
be identified and later filtered for Figure 11.
There are several methods for removing these noise elements. Among them include
using a bandpass filter to isolate the desired signal. This was the process performed for this
lab. Another way to remove this noise is to remove the extraneous data to better match
what the EEG is expected to look like. Both of these techniques have the advantage of
eliminating obvious noise artifacts above and below expected values, but falls short if the
noise is in a frequency range where the expected EEG signal is located. In order to help
minimize the level of biopotential noise, subjects are often asked to remain quiet and still
during the EEG recording. To truly eliminate the noise from both EOG and EMG sources,
the EEG, EOG and EMG data of a patient would be measured simultaneously. These
signals would then be subtracted from the EEG data to remove this noise. This method
does not account for alternative sources of noise, nor does it account for the cancellation
that may be present in a signal. Again, this method faces the risk of removing the wanted
EEG signals, mistaking them for noise. Additionally, this simultaneous measurement will
require a large amount of memory for a single recording.
Gold cup electrodes are frequently used for EEG recordings. They can also be
utilized to record EOG and EMG from facial muscles if desired. Using gold is specifically to
avoid any reaction between the metal and perspiration coming from the skin of the test
subject. This can be both a safety concern as well as a way to make sure the signal is not
affected by the sodium and potassium ions in the subjects perspiration. These electrodes
are smaller than typical metal disc electrodes and are shaped to look like a cup. The tiny
size of these electrodes makes it easier for placement on the head. Additionally, the smaller
size prevents the entrapment of hair beneath the electrode, which has the potential to
interfere with the readings. While the snap electrodes have an adhesive backing, cup
electrodes require a special conductive paste to adhere the electrodes to the skin. The
electrodes are filled with the conductive paste, pressed against the scalp, and then normally
taped to the head. Gauze is also common and may also be used to hold the electrodes in
place. For this lab gauze was utilized to help hold the gold cup electrodes in place.
When a more detailed and localized signal is desired from an EEG recording,
several methods can be used to improve the reading. As described earlier, there are four
types of brain waves that exist. When a specific wave is desired, a more detailed signal can
be gained by placing the electrodes on the scalp directly above where the wave typically
originates from. For example, if the alpha wave is the desired signal, electrodes can be
placed more densely on the scalp over the occipital lobe where the signal is expected to
originate. Additionally, it is possible to create a more dense net of EEG electrodes, such
that there are more electrodes spread across the scalp. This would allow for the more
accurate detection of the source of the waves, and may provide alternative signals when
noise is a complication. The analysis of the electrodes solely near the origin of the wave
was partially performed in this lab for Figures 5 and 6 where the channel selected for wave
analysis of alpha or beta corresponded to where that channel lead was placed on the
subject.
The signal can also be improved by having the subject perform the activities
normally correlated with the desired wave. Examples of this would be having the subject sit
quietly with their eyes closed to improve the alpha wave signal, having the subject engaged
in a mental activity for beta waves, having the subject recorded during times of stress for
theta waves, or recording during deep sleep for delta waves. The other methods to improve
the reading were previously discussed and included filtering and removal of noise and other
unwanted or unexpected signals. This was performed in the lab for Figures 5 and 6.
In this experiment, the electrodes used for the detection of alpha waves were located
above the occipital lobe (O1, O2). Detection of beta waves used the fronto-parietal
electrodes. No attempt was made in this lab to isolate theta or delta waves. However, theta
waves would use electrode placed on the temporal and parietal regions of the scalp. Delta
waves, associated with the deeper structures of the brain, would not have a particular
region where they would be recorded from. Instead, it would be better to use the electrodes
in areas where motion noise is not as likely to alter the reading; the frontal electrodes would
not be a good choice in this case. Since these are low frequencies waves, it is important
that low frequency noises are not a concern in the area of the scalp.


Conclusion:


This laboratory exercise consisted of first utilizing known information on how the
brain communicates using electrochemical potentials and then properly placing electrodes
using the 10-20 rule to record and process an EEG signal. After this signal was obtained, a
bandpass filter was applied to remove noise artifacts and later to isolate specific types of
brain waves (alpha, beta, theta, delta) based on their characteristic frequencies. Applying
the same techniques for noise removal and well chosen electrode placement, it was
possible to remove the majority of noise artifacts and isolate the desired waves using their
characteristic amplitudes and frequencies. While the noise may not have been completely
removed, its relative amplitude with respect to the rest of the EEG frequencies was much
less significant, allowing for a clear reading of the EEG signal.
From these activities we learned several important items about EEG signal
processing. One important note is that the EEG signals being recorded are actually the
summation of potentials, so small bits of noise normally expected can get amplified and
cause heavy signal artifacts. It was also learned that these sources of noise could be
removed like in other labs by using bandpass filters that removed artifacts outside of the
range that EEG signals normally come from. Since different brain waves originate from
different parts of the cerebral cortex as well as have differing characteristic frequencies, we
learned how to isolate one brain wave from the others. Once isolated, these brain wave
recordings can be used for diagnostic properties.