Session 9 - An Introduction to EEG and Event-Related Potentials

Question 1

Advantages

Answer 1

• high temporal resolution (ms range)
• non-invasive, few exclusion criteria
• portable (TU is developing one to walk with)
• relatively cheap
• all age groups

Question 2

Disavantages

Answer 2

• poor spatial resolution
• noisy signal

Question 3

Neural source of EEG

Answer 3

–> hypothesised
= postsynaptic potentials at apical dendrites of synchronised neocortical pyramidal cells

Dipole:
- net negativity at axon from presynaptic cells + dendrites
- net positivity at soma
–> generator of EEG = electrical dipoles

• chemical signal transmission at synapses (between neurons)
• electrical signal transmission within neurons

Cortex:
- in the stimulated region of the cortex: many tiny dipoles sum up to create equivalent current dipole

Question 4

Pyramidal cells

Answer 4

• create the postsynaptic potentials measured in EEG

Question 5

electrical signals can be measured at the scalp

Answer 5

• dipole orientation and structure determines the scalp topography

BUT:
- scalp topographies do not directly indicate the location of the source –> especially in the case of multiple dipoles

Question 6

The 10-20 system

Answer 6

• electrodes are usually mounted in electrode cap following the 10-20 system
  –> placement of electrodes is independent of head size and comparable across labs
• sectioning in frontal, temporal, central, parietal, and occipital

Question 7

Recording the EEG

Answer 7

we require:
- a stimulation computer
- EEG cap and participant
- filters & amplifiers
- digitisation computer

Question 8

Artefacts

Answer 8

• blinks
• movements/muscle activity
• eye movements
• sweat sway

Question 9

Data preprocessing

Answer 9

Steps (usually included):
- re-referencing
- blink correction
- filtering
- artefact rejection

main goal = clean the raw signal by
- substracting systematic artefacts (e.g blinks)
- filtering frequencies irrelevant to the research question
- rejecting data epochs with unsystematic artefacts (e.g. muscle activity, slow drifts)

• specific parameters of pre-processing depend on the analyses method (e.g. event-related potentials, time-frequency analyses)

Question 10

Quantifying the EEG signal

Answer 10

• ERP components –> time domain
• frequency analyses/time-frequency analyses –> frequency domain
• sources –> source domain

Other parameters and approaches include:
- use of single-trial data, e.g.
–> trial-by-trial covariation with reaction times
–> MVPA analyses of single trial activations
–> general problem: high levels of noise

Question 11

Event-related potentials

Answer 11

= activation related to an external (visual, auditory, tactile) stimulus

• ERPs inform about the time course of stimulus processing
• different components are linked to different processing steps - from sensory perception to higher-order cognition
• time course of experimental effects informs about neural architecture, e.g. top-down influence on sensory processing

Example: processing of emotional compared to neutral facial expressions
- presenting fearful, happy and neutral faces (~30-40 per condition)
- recording continuous EEG

Question 12

Example EEG Experiment

Answer 12

• faces are presented, e.g, x = neutral, o = positive
• data is saved continuously (500 Hz)
• event code (trigger) is saved at the time of stimulus onset

Question 13

Segementation

Answer 13

• after pre-processing: continuous data is cut into epochs around stimulus onset (identified by triggers), e.g. -100-1000ms
• triggers inform about onset and experimental conditions (e.g. a 1 is sent with every happy face, a 2 with every neutral face, etc)

Question 14

ERP-analysis

Answer 14

Averaging
- after segmentation –> (ideally) we have 30-40 segments for each experimental condition, cut around the onset of the words
- segments are then average for each condition in order to remove noise
–> random noise has a higher amplitude than activity related to stimulus processing
–> assumption that noise is random and cancels out through averaging

• averaging increses the signal to noise ratio

Disadvantage: loss of signal-trial information

Question 15

ERP components

Answer 15

• are linked to specific processing step of a stimulus
  –> P1 reflects first positive deflection (in visual processing in the extrastriate visual cortex)
  –> N170 reflects structural face encoding
• knowledge is based on combinations with other techniques like fMRI and PET, and confirmed by source analyses

Question 16

ERP: P1

Answer 16

• can be first positive amplitude or P100
• preak ~ 100ms after the onset of a visual stimulus
• reflects perceptual processing generated in the extrastriate visual cortex
• attention allocation increases P1 amplitudes

Question 17

ERP: C1

Answer 17

• first component (positive or negative)
  –> sounds stupid but was invented to describe the first amplitude change before P1/N1
• difficult to distinguish this from random activity

Question 17

Early components

Answer 17

• within ~200ms after stimulus onset
• mainly reflect sensory responses of the visual cortex and are highly influenced by physical stimulus properties
• can also be modulated by emotion, mood, attention, reward, etc

Question 18

Late components

Answer 18

• after 300ms
• relfect internal, higher-order processing
• not strongly dependent on physical stimulus properties
• influenced by task, strategy, emotionl processing, etc.

Question 19

Statistical analyses

Answer 19

• averages are calculated for every condition, separately for every subject
• ERP components of interest are quatified based on previous literature
  –> eg P1 component: activation from 100-200ms after stimulus onset at occipital electrodes
• values for each condition and each participant are entered into statistcial analyses (F-tests, t-tests, linear mixed models)

Question 20

Time frequency analyses

Answer 20

• EEG signal consists of a sum of simple sine waves of different frequencies
• frequency: number of events per unit time
• 1Hz: 1 cycle per second

Examples:
- alpha rhythm: 8-13Hz –> 8-13 oscillations per second
- sampling rate of EEG = 500Hz –> 500 data points per second, one sample every 2ms

Question 21

Frequencies: Fourier Transform

Answer 21

• EEG signal: sum of sine waves of different frequencies
• transformation from time domain to frequency domain by mathematical transformaiton (Fourier Transform)
  –> allows for quantifying the contributions of different rhythms to the EEG signal
  –> can be statistically compared across conditions

Problem:
- assumption of stationarity (the frequency composition does not change across time)

Question 22

Spontaneous rhythms in EEG

Answer 22

Gamma:
- 30 - 100Hz
- network binding of cognitive/motor function

Beta:
- 14 - 30Hz
- alert state

Alpha:
- 8 - 13Hz
- drowsiness

Theta:
- 5 - 7Hz
- sleep

Delta:
- 1 - 4 Hz
- deep sleep

• frequency bands are also related to specific cognitive functions, e.g. attention, perception, motor preparation, memory

Question 23

Source Analyses

Answer 23

Aim:
- localising the sources (dipole technique) of electro-cortical activity

Why not fMRI?
- because of EEG’s superior temporal resolution
- goal is to localise the sources of activation at specific processing steps

Source estimation involves two steps:
- creating a forward model of brain activation
- estimating sources using the forward model

Question 24

1. Step: forward model

Answer 24

Sources --> EEG Requires: - head model (anatomy) - laws of physics (how do electric fields spread through tissues?) --> simulate a dipole (source) and look at signal

Question 25

2. Step: inverse problem

Answer 25

EEG --> Sources - ill-posed problem: infinte number of solutions (how many sources are active?) - constraints on sources, geometry, etc. - source reconstructions are estimates