Flashcards in Week 4- EEG Deck (54):
What is the EEG?
- method of detecting neural activity by placing electrodes on the scalp
-electrodes pick up the electrical signals of the underlying neurons in the brain allowing inferences about neural activity to be made
It is completely non-invasive
Huge historical significance
In 1924, Hans Berger detected the first EEG signal with electrodes attached to the scalp of a human (results reported 1929)
Despite the observer not doing anything (eyes were closed), the electrical signal was not constant but varied with a characteristic frequency of 8-13 Hz.
Dubbed alpha rhythm
Used two electrodes, one attached to the front of the head and one to the rear. Recorded the potential (i.e voltage) difference between them
Initially, electrodes were silver wires placed under scalp
Later, sliver foil placed on the scalp
He claimed (correctly) that this fluctuation was caused by neural activity.
Initially, thought to be a crank, but findings finally accepted by scientific community in 1934
Although Berger generally regarded as father of EEG, his work was foreshadowed by Richard Caton
Caton placed electrodes directly on the brain of dogs and apes and had detected a small electrical impulses
Reported results in 1875, but was widely ignored.
How Does EEG Compare To fMRI?
Cost to run one subject for one hour
EEG $5 to $10
fMRI $500 to $1000
Cost of scanner and shielding
EEG $50,000 to $100,000
fMRI $1.5m to $3m
Temporal and Spatial Resolution compared to fMRI
Very high temporal resolution (~ ms)
Low spatial resolution (~ cm) (but depends on whether you use inverse dipole modeling or not)
Low temporal resolution (~ 5 s)
Moderate spatial resolution (~ 2 mm)
Origins of EEG Signals
The electrical signals generated by the neurons are what cause the electrical signals recorded in EEG experiments
EEG reflects the summation of electrical potentials generated by millions of neurons
A single neuron does not generate a large enough electrical potential to be detectable by an electrode placed on the scalp
EEG signals originate primarily from cerebral cortex.
Which is a sheet of neural matter on the outside of the brain
Origins of EEG Signals
Within the cerebral cortex, it is thought that the EEG signals are caused primarily by the currents in the apical dendrites of pyramidal neurons:
they tend to have synchronized activity
their dendrites are well aligned (so their electrical disturbances will sum together)
they are located near the scalp
Detecting activity from deeper sources (e.g. the midbrain) is possible but much more difficult.
Limitations of EEG
EEG signal is biased to signals generated in superficial layers of cerebral cortex on the gyri (ridges) directly abutting the skull
Signals in the sulci (troughs) harder to detect and may be masked by the signals from the gyri
The meninges, cerebrospinal fluid and skull “smear” the EEG signal, making it hard to localize the source
Primary limitation of EEG is its poor spatial resolution
Recording the Signal
Typically 20-256 electrodes are placed on the scalp using a recording cap
Usually a conducting gel is used between the electrode and the scalp to increase the EEG signal
Sometimes the scalp is lightly abraded to reduce impedance further
Sometimes individual electrodes fail to pick up a reliable EEG signal. These are removed from subsequent analysis
A popular method of arranging the electrodes (especially in clinical situations)
Electrodes A1 and A2 are placed on the O’s ears and used as reference electrodes
G is a ground electrode used by the amplification system to reduce electrical interference (don’t concern yourself with it at this stage – often not even needed).
A Quick Aside: Electrical Potential
Electrical potential cannot be measured at a single point
You cannot say “What is the voltage of that point”
Has to be measured relative to a reference point
In other words, you have to measure the voltage across two points
For example, one typically measures the electrical potential of a battery’s cathode relative to its anode
Naming of channels/signals
One cannot talk about the voltage of a single electrode.
Instead one can:
Compare the potential difference (aka voltage) between two electrodes, called “bipolar derivation” or “bipolar recordings”, e.g. F3-C3 signal/channel where F3 and C3 are the names of electrodes.
Compare the activity of an electrode relative to a common reference electrode, often placed on an earlobe, the nose, the mastoid (i.e. just behind the ear) or on the neck (e.g. A1 or A2 or and average of both)
Compare the activity of an electrode to the average activity of many electrodes (virtual reference)
EEG signals measured from the scalp have a typical amplitude of 10μV to 100μV
Consequently they need to be amplified, typically by a factor of 1,000 to 100,000
Signal is then (typically) digitalized. Typical sample frequency 256-512 Hz.
Signal band-pass filtered to remove the low (35-70Hz depending on the situation) frequencies.
Often “notch” filtered to remove artifact of power lines (in Australia 50Hz)
It might also be necessary to remove artifacts such as:
EKG (Elektrokardiogramm, i.e. cardiac) artifacts
EMG (Electromyography, i.e. muscle) artifacts
Glossokinetic artifact – caused by moving the tongue and consequently changing the electrical properties of the head
Eye-movement induced artifacts are the most common
Eye-Movement Artifact Removal
The artifact is first detected
Then it is determine which component of the raw signal is caused by this artifact
This component is then subtracted from the raw signal, resulting in an artifact-corrected signal
A similar technique can be used for other artifacts
- For example, one could measure EKG activity from an extremity to detect likely cardiac-induced artifacts
Analysis of EEG Signal
One can analyze both the periodic aspect of the EEG signal and the transients caused by the presentation of the stimuli
The periodic aspect of the signal is typically analyzed when the subject is not presented any particular stimulus (e.g. the subject is asleep or has her eyes closed)
Even in the absence of direct stimulation, characteristic frequencies are observed
The origin of these characteristic frequencies is still debated
Named Frequency Ranges
Delta (1-4 Hz). Typically generated by slow wave (deep)sleep.
Theta (4-7 Hz). Associated with drowsiness/light sleep
Alpha (8-12 Hz). Found when the eyes are closed and the O is relaxing/reflecting (but not asleep)
Beta (14-30 Hz). Found when O’s alert/working, eyes open
Gamma (> 30 Hz). Perhaps involved in short term memory?
EEG can be used to monitor depth of anesthesia
Especially useful when a muscle relaxant is administered to patient
One commercial system that does this is called the BIS (Bispectral Index) monitor.
Assigns a number (the BIS number) between 0-100 to the patient (0 = no EEG activity, 100 = wide awake).
Recommends anesthetised patient be kept in range 40-60
Precise algorithm used is proprietary, but essentially the higher the frequency of the EEG activity the more awake the patient is likely to be.
Other Clinical Uses of EEG
An adjunct test of brain death – a flat EEG is often good evidence that there is no neural activity
To monitor for non-convulsive seizures in intensive care
To distinguish epileptic seizures from psychogenic non-epileptic seizures (seizures of psychological origin)
To localize the brain area where an epileptic fit originates (typically done with iEEG)
To help patients interact with computers (e.g. EEG controlled wheelchair)
Intracranial electroencephalograms (iEEG) are recorded from grids of electrodes implanted in the cortex of patients
Typically used to (precisely) located the focus of epileptic seizures
More precise than EEG as it avoids the blurring effects of the skull and the associated structures
Localization of EEG Signal
Even in the non-clinical setting, it is often helpful to localize the EEG signal. This can be done by extrapolating between the electrodes to display (usually for a particular range of frequencies e.g. the alpha band) how the amplitude of the signal varies across the scalp.
Attempts to deduce the neuronal sources that give rise to distribution of electrical potential recorded on the scalp
Assumes that these sources can be thought of as dipoles that project positive and negative electric fields in opposite directions
The orientation, location and number of dipoles determines the electric fields recorded at the scalp
One has to discount the effect the brain and the skull have on electric fields – i.e. one has to solve the inverse problem
Only works, if neural sources are (in reality) well localized and small in number
There is still debate as to when this technique is appropriate
Event Related Potentials (ERP’s)
Now we consider the portion of the EEG signal caused by the presentation of a stimulus
Sometimes also called Evoked Potentials (EP’s)
Typically, present the stimulus multiple times and average the signals generated to obtain the ERP
Event Related Potentials (ERP’s) continued
The resultant evoked potential is typically described in terms of the polarities (P = positive, N = Negative) and the latencies of its components/waveforms
For example, the N1 component/waveform is the first negative-going potential
Actually, the N1 component is rather famous
Sometimes also referred to as N100 as it typically occurs 100 ms after the stimulus is presented
First studied by Pauline Davis in 1939.
Occurs when the stimulus is somewhat unexpected.
Strongest for unexpected stimuli, weakest for repetitive stimuli, might disappear completely when the observer generates the stimulus himself
Schafer & Marcus (1973)
Provided evidence that N100 waveform disappears if subject generates the stimulus
Indicates that the waveform is linked to novelty
Observers listed to short (1 ms), loud (80-db) clicks
Recorded voltage difference between two electrodes Cz-A. Data filtered 3-30 Hz
Schafer & Marcus (1973) continued
Self stimulation: Observers determine when the click occurred by randomly pushing a button (self-stimulation)
Periodic stimulation: Clicks occurred regularly every 2 seconds
Machine stimulation: Clicks are played to the observer pseudo-randomly
An endogneous potential linked not to the physical aspects of the stimulus but to the person’s reaction to the stimulus
First studied by Chapman & Bragdon (1964)
Usually occurs when something unexpected happens or when something expected does not happen.
The N2pc ERP component is a negative component that occurs approximately 200 ms after the presentation of the stimulus.
The letters “pc” stand for “posterior contralateral”. It is strongest over the posterior cortex contralateral to where the observer is attending.
Let us suppose two dots are simultaneously flashed before the subject, one on the left and the other on the right.
Furthermore, let us assume that we are recording two channels 01-G and 02-G
On each channel we will be able to measure an ERP
If the subject is attending to the left dot then the N2 component of the ERP will be greater on the 02-G channel than on the 01-G channel
When Subject Attends To The Left Dot
N2 component is strongest over right posterior cortex
Take Home Message- Schafeur
If the observer attends to the left side of space, the N2 component is larger over the right visual cortex. If the observer attends to the right side of space, the N2 component is larger over the left visual cortex.
In this way the N2pc component can be used to determine where an observer is attending
A lot of early ERP research investigated which manipulations would affect which components of ERP (e.g. if the observer expects to hear a click but then doesn’t hear a click, a large P300 component is observed).
But what does this tell us about how the brain work?
Why should we care?
It seems that people were studying ERP’s for the sake of studying ERP’s.
This was a common criticism of early ERP work, but more recently people have been able to use ERP’s to make inferences about how the brain works
visual search - hypothesis 1
: Search is serial.
Perhaps, because the target is so similar to the distractors, each item has to be attended to in turn to determine whether it is the target. Thus, search would be a serial process
visual search- hypothesis 2
Search is parallel
All items are attended simultaneously.
Woodman & Luck (1999)
used the N2pc component of the ERP to determine if search is serial or parallel
By monitoring the N2pc component, they could determine where the observer was attending
Remember, the N2pc component is stronger contralateral to the side to which the observer attends
What did woodman do?
Recorded the ERP from electrodes in approximately the 01 and 02 positions.
These recording were conducted relative to an unspecified ground electrode
By comparing the N2 component of these two electrodes they could determine which side the observer was attending.
Woodman & Luck (1999) continued
For example, if the observer attends to an item in the left hemifield, the N2pc component will be stronger over the right visual cortex.
Woodman & Luck used this fact to determine where observers were deploying their attention during a visual search task.
They could literally watch the observers shift their attention from one item to another, thereby confirming the serial search hypothesis.
Trial started with the presentation of a fixation cross, which remained visible throughout the entire trial. The observer had to maintain fixation on the fixation cross throughout the trial.
Then an array of squares were presented – all with a gap in one side.
The Stimulus continued
4 of these squares were coloured, 1 red, 1 green, 1 blue 1 violet (the remaining squares were black). The task was to look for the color square with the gap to the left.
Each observer was told that 75% of the time the target (i.e. the square with the gap to the left) would be a particular colour (e.g. blue) and 25% of the time it would be another color (e.g. red).
We shall call these colors C75 and C25
C75 and C25 were always in opposite hemifields
C75 is the colour of the item that 75% of time will be the target (i.e. be the square with the gap on the left).
C25 is the colour of the item that 25% of time will be the target
C75 and C25 are always in opposite hemifields.
If search is serial, then the observer will usually attend to the C75 item first.
Thus, we know where the observer will usually first attend.
if C75 Item Is The Target
Recall, N2pc is always strongest contralateral to the side to which attention deployed
Attention (almost) always deployed first to C75 item.
In a trial where C75 actually is the target then attention just stays on C75 item.
Consequently, N2pc component strongest contralateral to C75 item (i.e. dotted line above solid line).
Suppose C75 is blue square and C25 is red square
Then dotted line is left N2pc activity and solid line is right N2pc activity
Now Suppose C25 Item Is The Target
Again, attention first deployed to C75 item.
Consequently, dotted line is initially above solid line.
But now C75 is not the target so attention has to be switched to C25 which is in the other hemifield
Consequently, attention switches to the other hemifield and solid line rises above dotted line
In Other Words…woodman
Because the dotted line crossed over the solid line…
…this shows that attention crossed over from one hemifield to the other.
This directly shows attention switching from one item to another!
in more words- woodman
Event 1: Observer attends to the C75 item and this causes the N2pc signal to be more negative contralateral to the C75 item – shown by the dotted line being more negative than the solid line at approximately 250ms.
Event 2: Observer then attends to the C25 item thereby creating another N2pc signal, this one more negative ipsilateral to the C75 item (because the C25 item is in the opposite hemifield to the C75 item) – shown by the solid line being more negative than the dotted line at approximately 350ms.
This is why the dotted line cross the solid line.
The fact that the dotted line crosses the solid line shows that attention switched from one hemifield to the other
In Yet More Words…
Note that these two events each create an ERP and that the ERPs are superimposed on each other, with the second ERP delayed relative to the first ERP by about 100ms.
Thus, the arrow at 250ms corresponds to the N2pc component of the first event (i.e. the N2pc component of the first ERP).
The arrow at 350ms corresponds to the N2pc component of the second event (i.e. the N2pc component of the second ERP).
Conversely, when the C75 item was the target, attention went to this item and just stayed there.
Thus, there was only one event.
Woodman & Luck (1999) findings
therefore succeeded in watching attention switch from one item to another in a visual search experiment.
This was clear evidence that visual search occurred via a serial mechanism in which each item was attended to in turn.
In other words, at least for their stimulus, their data supported the serial hypothesis.
Notice, how we had to use EEG to do this experiment. We could not have verified the serial hypothesis directly from behavioral data.
We needed a method that would allow us watch where attention was deployed.
how does searching in memory for an item occur?
Two famous theories of attention make contradictory predictions:
The Theory of Biased Competition (Desimone & Duncan, 1995), predicts that any item held in memory will automatically attract attention, regardless of whether it is the target of your search.
Conversely, the Neural Theory of Visual Attention (Bundesen et al., 2005) posits that an item in memory will only attract attention when it is the target of your search
Carlisle & Woodman (2011) what they did?
-Target is the Landolt-square with gap up/down.
Subject has to find the target and report direction of gap (up or down).
-Subject then had to report whether memory test item matched previous memory item.
-Theory of Biased Competition predicts attention initially directed to memory-match distractor.
-Neural Theory of Visual Attention predicts attention not drawn to the memory-match distractor.
-Use EEG to distinguish between the two.
Predictions Theory of Biased Competition
-When memory-match distractor absent, attention will be drawn only to the target. Thus, N2pc will be highest contralateral to the target.
-When the memory-match distractor is on the same side as the target, attention will initially be drawn to the memory-match distractor and then will move onto the target.
-When the memory-match distractor is on the opposite side to the target, attention is initially drawn to the memory-match distractor and then moves onto the target. Thus, N2pc is initially greater ipsilateral to target (when attention is on memory-match distractor) then greater contralateral to target (when attention on target).