IVE 2024 Short Course - Lecture 8 - Electroencephalography (EEG) Basics

Electroencephalography (EEG) basics
IVE Winter School 2024
Ina Bornkessel-Schlesewsky, Matthias Schlesewsky
July 16th 2024
1

Physiological basis of the EEG
The EEG reflects summed dipole
moments, i.e. voltage differences
between higher and lower cortical layers
originating primarily between pyramidal
cells. These result from synchronous
postsynaptic activity.
4
Bornkessel-Schlesewsky
&
Schlesewsky
(2009)
IBS & MS - IVE Winter School 2024

Distance between electrode and
neural activity 5
Rösler (2005)
the distance between the
source of the current and
the electrode determines
scalp distribution
under certain circumstances,
the potentials of distinct
sources may merge at the
surface

The inverse problem
• Scalp-recorded EEG activity does not
allow any unique conclusions as to its
underlying sources
• There are mathematical models to
address the inverse problem, but all
solutions remain approximations!
6 IBS & MS - IVE Winter School 2024

EEG and different states of
consciousness
EEG activity occurs in different frequency
ranges
7
Gazzaniga
et
al.
(2014)

Linking EEG activity to
cognition
The human EEG was discovered by Hans
Berger (1873-1941), a German psychiatrist
1929: Über das Elektrenkephalogramm
des Menschen
(On the human electroencephalogram)
8
http://upload.wikimedia.org/wikipedia/commons/6/69/
HansBerger_Univ_Jena.jpeg

Berger’s EEG lab
9
Berger (1929)

An early EEG trace
10
Berger (1929)
EEG
ECG
time
(0.1s)
Berger applied electrodes made of silver foil
to the scalp. Today, most EEG labs use
electrodes made of silver/silver chloride ...

The first sleep EEG
a = deep sleep;
b = after waking from two hours of sleep
THE FIRST SLEEP-EEG
a = deep sleep & b = after waking from two hours of sleep
(top line = EEG; bottom line = metronome for timekeeping)
Berger (1929)

EEG & chloroform
• same participant without (top) and with
(bottom) chloroform
EEG & CHLOROFORM
same participant without (top) and with (bottom) chlorof
first line = EEG, second line = EKG, third line = metrono
Berger (1929)

EEG & problem solving
• Berger’s daughter’s EEG while solving a
maths problem
• top: EEG at rest (alpha waves, ~10 Hz)
• middle: during the task (beta waves; faster,
lower amplitudes)
• bottom: transition from task to rest
EEG
EEG from
Figu
Figure 2
Figure 3
Rest: al
around
Task: beta
Berger (1929)

Berger’s achievements
• Discovered the α (~8-12 Hz) and β (~12-30 Hz) rhythms
• Observed changes in the rhythmical activity of the EEG depending on changes in
cognitive state
• α decreases during problem solving (e.g. mental arithmetic)
• increases again during relaxed wakefulness
14

Event-related brain potentials
(ERPs)
&
Schlesewsky
(2009)
ERPs are small changes in the spontaneous
electrical activity of the brain that are time-
locked to certain sensory or cognitive events
15

ERPs
16
• Very small potential changes (in
cognition, often between
approximately ca. 2-8 μV) in
comparison to the background
activity (often in the range of
100s of μV)
• Low signal-to-noise ratio
• present a larger number of
stimuli of each type (approx.
30-40)
• average over a number of
participants (approx. 20-24) Kutas & Federmeier (2000)
N400

ERPs
Kutas & Federmeier (2000)
17
• An ERP component is an ERP
response that occurs with a
certain:
• latency (time after stimulus onset)
• polarity (negative or positive
relative to a control)
• topography (which electrodes?)
• Typical nomenclature: N (negativity)
or P (positivity) + peak latency, e.g.
N400
• Instances of a component with
different amplitudes (“strengths”):
effects
N400

Selected ERP components

The P300 (P3)
19
The first “endogenous” ERP component to
be discovered (tied to internal processing
rather than sensory stimuli)
Sutton et al. (1965, Science)

The P300 (P3)
20
• Aprominent positive deflection after approx. 300
ms (topography and latency vary according to
various parameters) Observable in response to
salient (e.g. unexpected — oddball), task-relevant
stimuli
• need not be unexpected; also found for self-
relevant stimuli (one’s own name)
• “motivationally significant stimuli”
• Longer latency for more complex stimuli,
amplitude varies e.g. with attention
• For discussion see Picton (1992, J Clin
Neurophys 9, 456-479), Verleger (1988, Behav
Brain Sci 11, 343-427), Nieuwenhuis et al. (2005,
Psych Bulletin) IBS & MS - IVE Winter School 2024

The first language-related ERP component
Kutas & Hillyard (1980, Science)

Measuring prediction errors
in the brain: EEG
the Mismatch
Negativity (MMN)
Oddball paradigm:
standard deviant
MMN:
early negativity
(between ~100 and
200 ms post stimulus
onset) for deviants vs.
standards; increasing
amplitude with
increasing physical
deviation between
standard and deviant
thought to re
fl
ect a
mismatch with a
transient sensory
memory trace Duncan et al. (2009, Clin
Neurophysiol)

e.g. to measure the e
ff
ects of
predictive cues in spatial augmented
reality (SAR)
Vollmer et al. (2023, IEEE:TVCG)
An oddball task as a secondary
task to measure “cognitive load”

Scalp EEG / ERPs: data acquisition

Electrode placement
&
Schlesewsky
(2009)
extended 20-20 system
26
Electrooculogram
(EOG) monitors
vertical (V) and
horizontal (H) eye
movements
Reference electrode(s)

The reference electrode
• The EEG is measured as a voltage difference between two electrodes, e.g. between the
electrode of interest and a reference electrode
• The choice of reference electrode determines the resulting signal
• Options:
27
common reference
average reference
active reference
inactive reference
(e.g. mastoid, nose)

The reference electrode
• Common choice in experiments on cognition:
• left or right mastoid
• + re-referencing to the average of both mastoids offline to avoid topographical
distortions
28

Why the choice of reference is
important
• oddball paradigm; active common
reference (close to PZ)
29

Why the choice of reference is
important
• oddball paradigm; same data
rereferenced to the left mastoid

31
Scalp EEG / ERPs: data analysis

Raw EEG data
32

Steps in data analysis
• Preprocessing
• Re-referencing
• Raw data filtering
• Artefact rejection (automatic + manual)
• Averaging
• single subject average (per electrode, condition and time window)
• grand average (average over single subject averages)
• Statistical analysis
33

Why filter?
34
before after

Filtering
• High pass
• removes frequencies lower than a specified frequency
• good for removing signal drifts from raw data (see last slide)
• generally between 0.1 and 0.3 Hz (not higher!!)
• must be applied to raw (not epoched) data!
• Low pass
• removes frequencies higher than a specified frequency
• good for “smoothing” averages for data presentation (typically 8-10 Hz)
35

Filtering
• Band pass
• combination of high and low pass filters, i.e. selects a frequency band
• Notch
• removes a specific frequency range from the data
• most common application: removal of 50 or 60 Hz (mains power frequency)
36

Artefacts I: ECG virtually eliminated by rereferencing
37

Artefacts II: eye movements
(blinks)
must either be excluded or corrected
38

(saccades)
not as disruptive as blinks
39

(saccades)
not as disruptive as blinks
40

Averaging
• Single subject average (per electrode, condition and time window)
• segments the raw data into stimulus-related epochs (length depends on question, but
usually between 800 and 2000 ms)
• options:
• absolute values
• relative to a baseline (e.g. -200-0 ms “pre-stimulus” or 0-100 within stimulus)
• excludes trials that contain artefacts (optionally: trials for which the control task was not
performed correctly)
• excludes noise and background activity
41

Averaging single subject average
42

Averaging II
• Grand average
• average of the single subject averages
• again per electrode, time window and condition
43

Averaging grand average
44

Averaging considerations
• Single-subject averages are (mostly) not interpretable
• signal-to-noise ratio too poor
• only evaluate data quality (no. of trials, ocular or movement artefacts, problems with
single electrodes, drifts ...)
• Grand averages
• the fact that a particular participant does not show the expected effect (or the effect
observed in the grand average) is not a sufficient reason for exclusion from the data
analysis!
45

Statistical analysis
46
• Average amplitude (in a specific time window)
• repeated-measures ANOVAs including
condition factors and topographical factors
• topographical factors: regions of interest
(ROIs)
• Peak-to-peak analysis
• amplitude differences may lead to shifts in
the time window
• analysis of the difference between the last
“common” peak and the critical peak
example ROIs

And always keep in mind
47
• ERPs are relative measures between a critical
condition and a control condition
• Absolute potential shifts cannot be
interpreted!!
two negative deflections

The “finished product”
48
• An ERP component that can be described in
terms of
• latency (time to component onset/peak
after critical stimulus onset)
• polarity (negative or positive deflection
relative to control)
• topography (electrode positions (ROIs) at
which the effect is observable
• (amplitude - “strength” of the effect)
• Nomenclature typically reflects polarity and
latency (e.g. N400)

Component versus effect
49
• An ERP component is typically associated
with a functional interpretation
• Components may also appear in control
conditions (e.g. N400 for every word)
• Differences between two instances of the
same component (and more generally
between two conditions): ERP effects

An alternative approach:
time-frequency analyses
Recall Berger’s (1929) observation:
frequency characteristics of the
human EEG can re
fl
ect higher
cognitive processes
50
-
+
theta
alpha 1
alpha 2
...

Classic EEG frequency bands
• Delta: 0.5 - 4 Hz
• Theta: 4 - 8 Hz
• Alpha: 8 - 12 Hz
• Beta: 12 - 30 Hz
• Gamma: > 30 Hz
51
NB: individual differences apply!

EEG frequency and higher cognition
• Independently of (and in parallel to) ERP-based research, a frequency-based research
tradition has identi
fi
ed correlates of di
ff
erent cognitive domains / processes within
di
ff
erent frequency bands, e.g.:
• alpha band
• attention (Ray & Cole, 1985, Science, 228, 750-752)
• memory performance (Klimesch, 1997, J Psychophysiol, 26, 319-340)
• intelligence (Doppelmayr et al., 2002, Intelligence, 30, 289-302)
• theta band
• episodic memory (Miller, 1991. Cortico-hippocampal interplay ... Berlin: Springer.)
52

Frequency bands:
The measures
53
Induced activity
(jitter in latency)
Evoked activity
(
fi
xed latency)
Averaged evoked potential
• evoked versus induced activity
IBS & MS - IVE Winter
School 2024

Isolating frequencies of
interest:
fi
ltering
• Filters remove certain frequencies from the EEG
• Di
ff
erent types of
fi
lters:
• High pass
• removes frequencies lower than a speci
fi
ed
frequency
• Low pass
• removes frequencies higher than a speci
fi
ed
frequency
• Band pass
• combination of high and low pass
fi
lters, i.e. selects
a frequency band
• Notch
• removes a speci
fi
c frequency range from the data
(most common application: removal of 50 or 60 Hz
mains power frequency)
54
Example from Widmann et al. (2014)

Event-related
(de-)synchronisation (ERD/S)
• Application of a bandpass-
fi
lter (frequency
band of interest) to each trial
• Squaring of each amplitude sample to
obtain power values
• Averaging over all trials of interest
• Conversion to relative power by de
fi
ning
the power of a reference interval as 100%
=> “event-related”
• ERBP (event-related band power): z-
transformed ERD

Frequency bands: The measures
56
Relation between ERPs and
measures in the frequency domain

Time-frequency measures of
performance
in complex operational environments
59
Cross et al. (2022, Scienti
fi
c Reports)

EEG analysis
60
• We use MNE (open-source Python package)
• https://mne.tools/stable/install/index.html
• Useful tutorials:
• Overview: https://mne.tools/stable/
auto_tutorials/intro/10_overview.html
• ERPs: https://mne.tools/stable/
auto_tutorials/evoked/30_eeg_erp.html
• Time frequency: https://mne.tools/stable/
auto_tutorials/time-freq/
20_sensors_time_frequency.html

IVE 2024 Short Course - Lecture 8 - Electroencephalography (EEG) Basics

More Related Content

Similar to IVE 2024 Short Course - Lecture 8 - Electroencephalography (EEG) Basics (20)

More from Mark Billinghurst (20)

Recently uploaded (20)

IVE 2024 Short Course - Lecture 8 - Electroencephalography (EEG) Basics