EEG Flashcards
what did Richard canon find out? and when?
Electrical phenomena in
animal’s brains (1875)
what did Adolf beck find out? and when?
Recordings of oscillatory
electrical brain activity in
response to different types of stimulation (visual, auditory) (1890)
who discovered Electrical phenomena in animal’s brains?
Richard canon
who discovered Recordings of oscillatory electrical brain activity in response to different types of stimulation (visual, auditory)
Adolf beck
who recorded the first human electroencephalogram? and when?
Hans berger 1929
who reported the first alpha waves? and what else did they report?
Hans berger
First report of alpha waves (~8 12 Hz) and their
suppression when the subject opens the eyes
electric
changes in membrane potential
chemical
synaptic transmission
intracranial measurements
- intraparenchimal recordings
- electrocorticography (EcoG)
extra cranial measurements
- magnetoencephalography (MEG)
- electroencephalography (EEG)
how is the temporal resolution of EEG?
excellent
on what time scale can EEG measure?
Measuring electrical (or magnetic) potential on a sub
millisecond scale
how is spatial resolution in EEG? and why?
low
-> because of volume conduction
-> difficulty recording activity from deep sources
->inverse problem
what is volume conduction?
The spreading of electrical
current from the source to the receiver while crossing
different media
is volume conduction an issue with MEG?
no
what is difficult to record with EEG and why?
Difficult to record activity
from deep sources
> Electrical field spreads as a
function of distance
> The activity from neural
populations with different
orientations cancels out
what is the inverse problem?
Estimate the location of brain activity given the pattern of activity measured on the scalp
> ill-posed problem: impossible to solve perfectly
what are the advantages of EEG? (4)
- High temporal resolution
(milliseconds) - Directly measures electrical activity of the brain
- (relatively) mobile and non
invasive - (relatively) cheap
what are disadvantages of EEG? (3)
- Poor spatial resolution
- Very weak signal (small
signal to noise ratio) - Difficult to measure deep
structures
where is electric potential measured from?
Electric potential is measured from an array of electrodes placed
on the scalp with respect to a reference point (different possible locations, but mostly also on the scalp)
how is the position of the electrodes on the scalp?
standardized
what is the system for the electrode placement?
> 10- 20 system
what is the most common number of electrodes? and why?
64 -> good trade off
how many electrodes can be used?
max 256
what are pro/cons of more electrodes?
- more spatial accuracy
- more time to prepare
where are additional electrodes placed and why?
Additional electrodes placed on the face to record EYE
MOVEMENTS
how are the additional electrodes placed?
> Horizontal electrooculogram
HEOG ): to the corner of the eyes
Vertical electrooculogram
VEOG ): above and below the eye
what is the equipment? (9)
- cap
- electrodes bundles
- gel
- flat electrodes
- amplifier
- battery pack
- trigger box
- computer to record EEG data + software
- computer to administer the task (if needed)
what is the EEG procedure? (3 steps?
- step 1: place cap on the participant’s head
- step 2: add gel in each hole of the cap
- step 3: insert electrodes
what is the goal of data cleaning?
Remove all the noise and keep only the signal coming from the brain
what are you filtering out?
- Low frequencies (drifts)
- High frequencies (muscle contraction)
- Line noise (50 Hz)
what are parts of preprocessing?
- filtering
- referencing
- interpolarting bad channels
what is referencing?
Change in voltage with respect to this reference (external electrodes, average of all channels, one specific channel)
what is interpolating bad channels?
Correcting data from one electrode based on the average of its surroundings
what is epoching?
epoching is a procedure in which specific time-windows are cut from the continuous EEG signal
what is the independent component analysis (ICA)?
Decomposes the signal in independent parts explaining a portion of variance
what does clean data look like?
The largest sources of noise have been removed, the signal from the brain is retained
what is univariate?
- (activation-based)
- Comparing levels of activation across conditions, usually averaging across
recording units
what is multivariate?
- (information based)
- How spatially distributed activity patterns contain task relevant information
examples of multivariate
-Decoding (machine learning)
- Representational similarity analysis
- ICA, PCA..
what does each portion of recorded data contain? (ERPs)
- signal and noise
what cancels out the noise?
averaging across trials
difference between signal and noise?
signal: similar in each trial
noise: randomly fluctuates across trials
what can you see after many many trials?
Scalp recorded systematic neural signal generated by a specific neuroanatomical module when a specific computation is performed
what did hallowed and Paulin Davis record? and when?
Recordings of sensory ERPs (1939)
who recorder sensory ERPs?
Hallowell and Pauline Davis (1939)
who recorded the first cognitive ERP? and when?
Grey Walter (1964)
what did grey walker find?
First cognitive ERP
-> Contingent negative variation
what is contingent negative variation (CNV)?
- Condition A (No-Go):
Passive viewing - Condition B (Go):
Press a button upon
detection of the target
stimulus
what does contingent negative variation reflect?
Reflecting preparation for
the upcoming target
sensory & perceptual ERP components (2)
- Visual P1
- N170
what is P1? and what does it reflect?
Large positive component,
most prominent at lateral
occipital electrodes, peaking roughly 100 ms post stimulus .
Reflects visual processing
what is N170?
Lateralized posterior negative peak associated
with face processing
what is the motor ERP component?
Readiness potential
Negative deflection preceding a movement , maximal above (contralateral) motor areas.
where is P1 most prominent?
most prominent at lateral
occipital electrodes
when does P1 peak?
peaking roughly 100 ms
post stimulus
cognitive ERP components (3)
- P300 (and subcomponents)
- mismatch negativity
- error-related negativity
what is P300 sensitive to?
Sensitive to target probability , peaks
over frontal electrodes
what is mismatch negativity?
Negative wave maximal at midline scalp sites, ~ 200 ms
post a deviant stimulus
where does mismatch negativity occur?
at midline scalp sites
when does mismatch negativity arise?
scalp sites, ~ 200 ms
post a deviant stimulus
what is error-related negativity?
Negative deflection at
frontal and central sites
with onset immediately following an erroneous
response
where is error-related negativity?
Negative deflection at
frontal and central sites
when is error-related negativity?
with onset immediately following an erroneous
response
what are oscillations?
The recorded data can be decomposed in signal (s) and noise
what is the phase (oscillations)?
the angle at a given time
what is the frequency?
the cycles per second (width)
what is the amplitude?
how low/high the oscillation is
what does the highest frequency you can look at depend on?
depends on your sampling rate
frequency bands (5)
- delta
- theta
- alpha
- beta
- gamma
what is Delta associated with?
Associated with some sleep stages
Hz of delta
1-4 Hz
what is theta associated with?
Hippocampal theta associated with learning and
memory, midfrontal theta with cognitive control
Hz of theta
4- 8 Hz
what is alpha associated with?
Thought to inhibit the cortex not in use (i.e.
gatekeeping), for example in attention
Hz of alpha
8 - 12 Hz
what is beta associated with?
Linked to waking consciousness, active
concentration, motor processing
Hz of beta
13 - 30 Hz
what is gamma associated with?
Involved in conscious perception, WM and
attention
Hz of Gamma
30 - 70 Hz
where is alpa amplitude suppressed?
Alpha amplitude is suppressed in parietaloccipital
electrodes contralateral to the attended location
effects of cue validity
contralateral - ipsilateral
60% - 80% - 100&
what is machine learning?
Tries to learn how the brain
responds to different
conditions
what is the goal of machine learning?
The goal is to
correctly predict to what
condition new, unseen data
belong to
3 steps of decoding in ML
1) Train classifier on a
subset of labeled data
2) Test classifier on
new unlabeled data
3) repeat many times
very large multidimensional data space
Need to be careful of
multiple comparisons
problem and implement
specific statistical measures
EEG is a … technique to …
Great technique to
study fast cognitive
dynamics, not ideal for
‘where’ questions
Rich, complex, and
multidimensional data that
can be analyzed in many ….
ways depending on the
research question. Need to
be careful with statistical
testing