EEG Flashcards
Electroencephalogram EEG
Measured by placing a cap with electrodes in the head that measure electrical brain activity
-> electricity from postsynaptic potentials (electricity changes when neurtransmitters connect to the new neuron)
- postsynaptic potentials take longer and can travel to the skull. Action potentials are too quick
Dipoles
Postsynaptic potentials create electric dipoles at the pyramidal cells
- plus and minus side cause a magnetic field (like a battery)
- depend on where the interaction is taking place and if the neurotransmitter is excitory or inhibitory
From neuron to electromagnetic field
Excitatory post synaptic potential causes an action potential
Excitatory neurotransmission
Excitatory neurotransmitter will lead to the cell becoming more positive (and reach action potential). The outside of the cell will become negative relative to the cell itself, which together they create a dipole, flowing from + to -
Inhibitory neurotransmission
Inhibitory neurotransmitter will lead to the cell becoming more negative (away from action potential). The outside of the cell will become positive relative to the cell itself, which causes a dipole away from the cellbody.
- reverse polarity of the voltage we record from the scalp
From dipoles to EEG
If the body is pos and the axon/dendrites are neg, then the EEG will dip into negative
If the body is negative and the axon/dendrites are pos, then the EEG will dip into the positive
-> so which ever is closest to the skull wil show on the EEG (positive or negative)
Timing of postsynaptic potentials for EEG
When they fire at the same time, it causes a stronger signal
Orientation of neurons and postsynaptic potentials in EEG
If they face opposite ways, they will cancel out eachother (positive + negative = neutral)
- if they face the same way, signal will become stronger
- in subcortical structures it will happen that they cancel each other out. We can only really measure the outskirts of the brain
What is measured by EEG
Post synaptical potentials at the apical dendritic trees of pyramidal cells
–> signal needs to be strong
- all at the same time
- all oriented the same way
- pyramidal cells have both of this in the cortex
Experimental EEG set-up
- One computer that is showing the stimuli
- One computer that measures the brain activity
-> both of these have to be connected to each other, so we know what is being shown at what point in the EEG - electrodes are connected via filters and amplifiers to the EEG computer
How do the electrodes of EEG measure/work
You measure the difference between two electrodes
- one on the scalp (active)
- other is a reference (passive -> for example earlobes)
-> result in rhythmic fluctuations in voltage
So it measures the difference between the two and that is what shows up
Reference and ground electrode locations
- tip of nose
- nasal cavity
- earlobe
- mastoid (behind the ear)
Fundamental principle of reference electrodes
Always think of ERPs as a difference between the active and reference sites
- not biased to one hemisphere (both sides if it is the ears)
- easy to attach
- not distracting
- frequently used by other investigators so the waveforms are comparable
Active, reference and ground sensors
Voltage between Active - Ground
Voltage between Reference - Ground
(A - G) - (R - G) = A - R indirectly
–> like ground doesn’t exist, but it will take out noise
Electrode placement
Fp: Frontal Pole
F: Frontal
C: Central
P: Parietal
O: Occipital
T: Temporal
Left are odd numbers, right are even. Midline is zero (-> the more to the midline, the lower the number)
- International 10-20 system
- International 10-10 system
The international 10-20 system
Refers to the distance between adjacent electrodes in percentage, which makes sure it is alwaysof equal proportion
- Nasion, Inion, Vertex en Preauricular point
- 10-10 is with 10% intervals
Electrodes
Contact point is often silver or silver chloride (Ag or AgCl)
- high resistance due to skull, a conductive gel lowers the resistance
- ground electrode reduces the environmental noise
- reference electrode provides biological baseline
- EOG (eyes), ECG (heart) and EMG (muscles) can be used simultaneously.
Electrooculography (EOG)
Can be places around the eye (vertical up or down (on the left side); horizontal left or right) to measure eye activity
EEG amplifier
Is a middle step between the input from the electrodes and the output on the screen. Makes the waves more noticible to be able to read the EEG better.
Analogue to digital conversion EEG
Input is in voltage, but the computer can’t do anything with that. Needs to be translated to something the computer can use; one’s and zero’s. We do that with sample rate
Sample rate
Time series will represent the voltage values
- sampling frequency: the rate of digitization in Hertz (Hz)
- sample rate of 500 Hz: 500 x per second = each 2 ms take one sample (per electrode)
Undersampling
Too little samples which causes an inaccurate representation in the digitalized version -> aliasing
- Nyquist-Shannon sampling theorem
Nyquist-Shannon sampling theorem
Sample rate should be at least 2x the fastest frequencies in the signal
-> fixes aliasing
Resolution
Amount if information in each sample -> the more the better
EEG: different frequencies
- gamma
- beta
- alfa
- theta
- delta
-> beschrijf hoe ze er ongeveer uit zijn, klein of groor, irregualr of niet etc. slide 48
Properties of the field signal
The field activity oscillates in time. thus the signal is represented as a time wave. AMplitude and frequency and phase are used to describe and analuze the signals
Event related potentials ERP’s
We want to go from raw data to an ERP
–> EEG changes that are time locked to sensory, motor or cognitive events that provide safe and noninvasive approach to study psychophysiological correlates of mental processes
How to make an ERP
You do an experiment and you categorize the output in your stimuli. You put them all together and combine them into an average. Now you have an average reaction to your stimuli and you can start the interpet the information.
Epoch
The different categories in ERP’s that you make
Importance of clean data
- noise in ERP
ERP’s are tiny (effects are less than a few millionth of a volt)
- So they are imbedded in noise
- averaging will reduce noise
–> Signal/Noise ratio: number of trials will increase the ratio
Grand Average waveform
In papers the EEGs that are shown are averages of lot’s of participant together
- analysis is still done on individual bases, but the combined imaging is better in showing the effects without noise
ERP’s are continuous
Allows to measure the brain processes that occur between the stimulus and the response instead of just measuring the response.
- P and N peaks, P is pos. N is neg. P1 is the first pos. peak that we see, the next one would be P2 etc.
Types of peaks in ERP’s
Generally early peaks are more sensory processing and later peaks are more cognitive and deeper processing
- peaks in sensory processing are not comparable (for example visual and audible)
- peaks in cognitive processing are comparable
Physical property sensitivity in ERPs
The C1, P1 and N1 waves will differ if there are different brightness level for example. More intense peaks
ERP parameters
- we can look at the peak
- we can look at the latency when after the stimulus will the peak appear)
- look at peak to peak
- measure the area of a peak
Topography of ERPs
All ERPs of the different sensor display on a head, so on the sensor locations there are ERPs displayed
–> the location of the ERP is not necessarily where the neural source is: when the source is at an angle, the pos and neg electricity can go different ways
Aural activity in topography
Activity will be shown on the sides and the top of the head, but the auditory system is on the side of the head
- it shows on the top, because the system is tilted and the waves point to the top of the head, so that is where they are being picked up
–> topography and neural activity are not always in the same location
Peaks are not components
Stimulus activated different areas that go in different directions, so for one ERP there could be an interaction with multiple sources. One ERP is not one source
–> ERPs are a combination of the neural sources
What is component
N1 is a component and then P1 is a component en then N2 is the next. You can look at these individually, but when you put them together you get the ERP
Inverse problem
It is very difficult to backtrack to where the source is. We only see the ERP output, but we still don’t know what happens where in the brain
Scalp topography
The ERP that are being shown can also be coloured, with a deeper color for intenser peaks
- ERP that look the same can have different topographies. It depends on the difference in the individual components.
- same ERP results, but different configurations.
Magnetoencephalography MEG
Measures magnetic activity instead of electrical activity in EEG –> this helpt locate the source of the activity
Comparisons EEG and MEG
- Electromagnetic field is a physical field produced by electrically charged objects
- it has properties of both electricity and magnetism
- electric and magnetic field are oriented perpendicular (staan haaks op elkaar)
Why are EEG and MEG similar and can we use them
- they have a common origin
- the spatio-temporal resolution is similar
- MEG: magnetic field permeates biologiscal tissue, fluid and air
–> less distortion and smearing out of the signal - MEG is better for localization, but costs a lot more
Why use EEG
- reaction is the final outcome
- EEG can track the time course wit millisec. precision
- EEG can inform about cognitive processes when there is no behavioral respons
Why not use fMRI instead of EEG
- ERPs allow us to measure the good stuff
–> everything in between the stimulus and the final reaction
Comparisons between EEG and fMRI
fMRI will show location and EEG will show the timing
N170
- face specific component
- respons more to faces than to non-faces
- occipital electrode sides
- somewhat stronger on right hemisphere
Mismatch Negativity MNN
MMN is evoked automatically by a change in a sequence of sounds
- even when doing something, your brain will catch this and show a mismatched negativity
ERP components related to language
-In N400 incongruent words in sentances created bigger activity
-In P600 grammatical errors cause bigger peaks
–> you can combine both as well
ERP related to emotion
Late positive potential (LPP) primarily reflects arousal and not valence (pleasure)
Steps in data analysis of ERP
- preprocessing
- main signal processing
- statistical testing
Preprocessing
- (re)reference of the EEG signals
- preprocessing reduces noise and formats EEG data for the man signal processing (filtering, correction for eye movement and artefact rejection)
- segmentation
Referencing
We use a reference electrode during the data collection
Rereferencing the reference
Cz - ((Lm + Rm)/2) –> Cz is sensor, Lm is left mastoid (ref.) and also for right mastoid
- can also do the average of all the electrodes together
- ERPs can look very different with different references
Examples of noise
Electric signals not from the brain
- EMG
- EOG
- ECG
- skin potentials
- respiration
- body motion
- environmental noise
- measurement noise, bad electrode
- artefact due to co-registration with fMRI and TMS
Noise reduction
- visual inspection and artefact detection and rejection
- topographic interpolation
- independent component analysis (ICA) for removing eye movement artefacts
- filtering
Automatic artefact rejection
Delete the segment from the data if some criterion is met, e.g. if amplitude of signal >200 microV
- the criterion is in the analysis, so therefor this will be done automatically
Artefact rejection
- trade-off between settings of artefact rejection and s/n ratio of ERP
- liberal artefact rejection
- conservative artefact rejection
Liberal artifact rejection
Many trials in the average but contaminated with artefacts
Conservative artefact rejection
Fewer trials in the average, though less artifacts
Topographic interpolation
- When one or a few electrodes are very noisy throughout the experiment
- replace the electrode with the mean of the surrounding electrodes (everything in circle around it in the topograph)
Ocular rejection
Ocular artefacts like blinking are visible in the EEG, these are rejected and taken out of the EEG
Ocular correction
No rejecting and taking out the ocular artefacts but correcting them –> independent component analysis ICA
Independent component analysis ICA
Make smaller groups of individual components –> unmixing
Take out the artefacts and remix the rest
Filtering
Filtering is removing high or low frequencies from the EEG signal (fast or slow moving data)
- remove non-neural voltages (50 Hz, movement of electrodes, muscular artifacts, sweating etc.)
High-cut or low-pass filter
Remove high frequency noise
-> cut the high freq. and let the low pass.
Low-cut/high-pass is the opposite
Band-reject/Notch filter
Filters out noise between the lower and upper thresholds
Segmentation
Every condition has a unique marker
- cut the EEG into segments (epochs) separately for both markers
- start and begin time epoch
Baseline correction
Separately per electrod:
- compute the average baseline activity
- substract the average baseline activity from each sample in the epoch
Peak detection
Look for the largest peak per participant for the epoch and you will find and average number
Mean activity
- mean activity over some period
- when large latency variation in peaks
- when there is no peak but more sustained activity
- when difference between conditions lies between peaks
Peak or Area export for SPSS
- but which peak at which electrodes should we test?
- A priori: based on predictions and literature
- see virtual ERP bootcamp for risk of ‘double dipping’
Dara driven appraoch for discovering differences between ERPs
- determine differences between two conditions by data driven approach
–> at which time at which electrodes do the ERPs significantly differ? - for each sample at each electrode: test the difference between two ERPs by t-tests
Problem with data driven appraoch
High type 1 error (false positives): accepting the alternative hypothesis when you should’t
- too many tests, so the chances are too high
- there are several statistical procedures to correct for type 1 error
Fourier transformation
Decompose the signal into its frequency components
- socillating signals like EEG consist of sine waves of different frequencies
Fourier transformation: find the sine waves that make up the original wave
Frontal alpha asymmetry FAA
FAA is an indicator of asymmetric brain activity in the frontal cortex, which refers to asymmetrical anterior activity in the alpha band
- alpha power is inversely related to regional brain activity, since you are drowsy when there is alpha
Different power strength left and right FAA
Take the average from both left and right
Right - Left
–> FAA = 0 = perfect symmetry
–> FAA = - = more right activation –> meer negatieve emoties
–> FAA = + = more left activation –> meer positieve emoties
What does FAA represent?
- valence model
- approach withdrawal model
Valence model
- negative vs. positive emotions
Greater left hemisphere activity (lower alpha power) is associated with positive emotios, whereas greater right hemisphere activity is associated with negative emotion
Approach-withdrawal model
Positive emotion often correlates to approach-related motivation with more left frontal brain activity, whereas negative emotion correlates to withdrawal-related motivation with more right frontal brain activity. Emotions with approach motivational tendencies are linked to a higher left frontal activty, whereas emotions with withdrawal motivational tendencies are linked to a higher right frontal activity.
Why Time fruquency analysis
- Fourier analysis of EEg data is done over quite a long period
- you lose the timing information and that is the whole point of EEG
- what if you want to know how the power of certain frequencies change over time
- do time-frequency analysis
Time frequency analysis
??
Why Multivariate pattern analysis MVPA
How does the brain differentiate two stimuli
- can we say what we are looking at based on the EEG data?
- compare the ERPs for both (old)
- MVPA (new)
Nog een keer naar MVPA kijken
laatste onderdeel
OP slide 190
Alle terminologie, handig om mee te leren
