Methods & Techniques in Neuroscience: EEG

HideShow resource information
  • Created by: CanveySam
  • Created on: 01-05-15 15:31

Why EEG?

  • Excellent time resolution

    • –  Cognitive, perceptual, linguistic, emotional and motor processes are fast and dynamic

    • –  For example, consider theta band (4-8 Hz), a ‘slow’ rhythm but quite ‘fast’ for our conscious experience

    • –  Or consider gamma (30-80 Hz)

  • Direct indicator of neuronal activity

  • Multidimensional (time, space, frequency, power, phase, connectivity etc)

  • Portability (observing brain in action)

  • Relatively inexpensive 

1 of 33

Why not EEG?

  • It is not well-suited for precise functional localization

  • It is not well-suited for measuring deep brain structures (e.g., putamen, thalamus, nucleus accumbens)

    Sub-optimal method: where in the brain does process X occur or is information Y stored

• It is also not very well-suited to study very slowly fluctuating process with uncertain and variable time course 

2 of 33

EEG basics

EEG reflects the differences of electrical potential over time, created by the current flows originating from neuronal populations 

Two types of neuronal electrical activity:

(i) Action potential (AP)

(ii) Postsynaptic potential (PSP) 

Synapses:

Chemical synapses cause local changes in postsynaptic membrane potentials, through neurotransmitters.

Information transmits with some delay on the order of a millisecond.

Besides chemical synapses there are electrical synapses, or gap junctions.

Ions flow directly through large channels into adjacent cells, with no time delay. 

3 of 33

Post-Synaptic Potential (PSP)

  • An electrical potential initiated at a postsynpatic site that can vary in amplitude and spreads passively across the cell membrane, decreasing in strength with time and distance

  • Generation of PSP

    • –  When AP reaches presynaptic axon end, a neurotransmitter is released into the synaptic cleft

    • –  The neurotransmitter binds to the receptor of the postsynaptic neuron by opening or closing an ion channels

    • –  This lead to a graded change in membrane potential

  • Two types of PSP
    – Excitatory PSP (for excitatory synapse) – Inhibitory PSP (for inhibitor synapse) 

4 of 33

Electrical signals

Electrical Signals are the Vocabulary of the Nervous System

Neurons perform information processing to integrate synaptic inputs.

A postsynaptic neuron will fire an action potential if a depolarization that exceeds threshold reaches its axon hillock.

Generally the combined effect of many excitatory synapses is required for a post-synaptic neuron to fire. 

5 of 33

Summations

Two Types of Summations

Spatial summation is the summing of potentials that come from different parts of the cell.

If the overall sum – of EPSPs and IPSPs – can depolarize the cell at the axon hillock, an action potential will occur.

Temporal summation is the summing of potentials that arrive at the axon hillock at different times.

The closer together in time that they arrive, the greater the summation and possibility of an action potential. 

EEG signals are primarily produced by summation of postsynaptic potentialsof millions of neurons 

6 of 33

What ISN'T EEG?

But what EEG is not

  • It cannot measure all neural events

  • It cannot measure individual molecular or synaptic events nor it can isolate events that are produced by a specific neurotransmitter or neuromodulator

  • It is less sensitive to deep brain structures

    – Field strength decreases exponentially with distance

    – Neuronal populations in deeper structures are not arranged in a geometrically parallel fashion

  • It is not very suitable to measure to very slow (< 0.1 Hz) or very high (> 100 Hz) fluctuations 

7 of 33

Sample questions

Sample MC Questions

EEG signals represent summation of

A. actionpotentials

B. post-synapticpotentials

C. a mixture of A and B

D. neither A nor B

Which filter is often used to remove line noise?
A. High pass filter

B. Band pass filter

C. Band stop filter

D. Low pass filter 

For an EEG signal with maximum frequency of 70 Hz, aliasing occurs when

A. fs=256Hz

B. fs=1024 Hz

C. fs=512Hz

D. fs=128 Hz 

Which EEG oscillation is most relevant for visual feature binding?

A. Alpha

B. Gamma

C. Beta

D. Delta

Which ERP component is associated with semantic violation in language?

A. MMN

B. P300

C. N400

D. N170 

Sample Brief Questions

(a) Describe briefly how post-synaptic potentials are generated.

(b) Also describe briefly the procedure by which the post-synaptic potentials are integrated 

8 of 33

Electrodes 1

Metal (conductive)

  • –  Ag/AgCl Electrodes (Silver electrodes with a thin coating of silver- chloride

  • –  Tin Electrodes

–  Goldcap Electrodes

The conductivity should be good between the electrode and the scalp

  • –  Electrode gel to reduce the impedance/resistance

  • –  Impedance below 5 Kilo Ohms

  • –  Scalp preparation (removal of dead skin cells)

    Active Electrodes

  • –  Integrated pre-amplifier

  • –  Faster preparation time 

9 of 33

Electrodes 2: Location and placement

Electrode Locations?

• International 10-20 Electrode Placement System

Electrode Naming and Placement

Fp = Frontal pole C = Central O = Occipital T = Temporal P = Parietal 

Odd ending: LH Even ending: RH

Larger the number, larger the distance from midline

Electrode Number? 

Rule of Thumb: Unless you expect precisely localized brain activity, 64 electrodes will be sufficient 

Traditional - 19

Standard - 32-64

High density - 128-256 (or more)

 - Pros: Better spatial sampling, Source reconstructions

  - Cons: Poorer signal quality, long prep time, electrolyte bridge

 

 

 

10 of 33

Electrodes 3: 3 step procedure

With active electrodes:

Place headcap

Apply electrode gel

Connect electrodes 

With passive electrodes:

Prepare skin and then as above

11 of 33

Amplifiers

  • The signal is amplified from a few μVolts to a few Volts.

  • The amplification is done by Differential Amplifiers

– Three electrodes:

Active Electrode (A) placed at the desired site
Reference Electrode (R) placed elsewhere on the scalp

Ground Electrode (G) placed elsewhere on the scalp/body

– AmplifiesAG–RG(whereasAG=A–G; RG=R–G)

Elimination of ambient noise

• Works best when impedances are same (low) for A and R

• Amplifier gain: 5-10 K
• Optimal gain depends on the 
input potential and output range 

12 of 33

Reference Site

  • Preferably a ‘neutral’ site (tip of the nose, the earlobes, the mastoids, the chin etc)

  • Three practical criteria
    – Choose a site that is convenient and comfortable
    – Choose a site that does not induce hemispheric bias – Choose a site used by other researchers in your field

  • Mostly used ‘neutral’ references: – average of two earlobes
    – average of two mastoids

  • Other referencing scheme:

    – Average of all electrodes

    – Current source density maps

    • Reference free method

    • Requires high density recording

    • Less accurate for boundary electrodes

    • Insensitive to deep sources 

13 of 33

Analog filters

Why filtering?

Avoiding ‘aliasing’ (sampling frequency is less than 2 x maximum frequency)

Reducing artefacts

Four types of filters:

  • Low pass - 100 Hz

• High pass - 0.5 Hz (or 0.1 Hz for slow brain responses)

• Band pass

• Band stop, Notch - 50 Hz (for removing power line noise; 60 Hz in USA)

Digitization (Analogue to Digital Conversion)

• 16/24 bit Resolution (216 or 16192 different voltage values can be coded by the ADC) 

  • Avoid aliasing 
  •  Sampling frequency (fs) should satisfy Nyquist Criterion fs > 2 fmax (fmax = max. frequency of interest) 
  • In practice, fs > 5fmax 

 

14 of 33

Artefact elimination

Two Kinds of Artifacts Elimination

Artifact Rejection
• Essentially a ‘signal detection problem’
• ‘Brute force approach’: Reject if over threshold (75-100
μV)

• artifacts usually have much larger amplitude
• Blink (Check vEOG, Topography, Polarity)
• Eye movement (Check hEOG, Step-like wave)
• Electrode shift (Shifting of potentials)
• Muscles (High frequency)
• Heart (Mostly in mastoid electrodes, Low frequency)

Problems

• Loss of significant portion of data
• Some participants are very prone to certain artifacts • Some tasks essentially call for artifacts 

Be careful with artefact rejection as too many artefacts can be lost; an alternative is artefact correction where arithmetic algorithms are used to CORRECT the data  for the artefacts. But be prudent!

15 of 33

Artefact correction

Philosophy: “Don’t throw the baby out with the dirty water, but clean the water and throw the dirt only!”

Simple methods

• Subtraction method (variance based)

• Filtering

Advanced methods
• Dipole/Source modeling procedures
Independent Component Analysis (ICA) 

Minimization of Artifacts

  • Electrical screening of the testing space (Faraday cage)

  • Careful instruction of partipants to minimize movement; blink pauses

  • Ensuring the participants in relaxed condition (to reduce muscle activity)

  • Careful electrode application to minimize impedance

  • Maintaining cool temperature and low himidity level inside lab (to reduce slow drift)

  • Filtering (e.g., high-pass filter to remove slow-shifts [i.e., low-frequency fluctuations in the EEG], as well as low- pass filter to avoid aliasing=bandpass filter) 

16 of 33

EEG Oscillations

Standard Frequency Bands:

  • Delta: < 4 Hz

  • Theta:4–7Hz

  • Alpha: 8 - 14 Hz

  • Beta:15-30Hz

  • Gamma > 30 Hz

This division is arbitrary, mostly by convention 

17 of 33

Applications of spontaneous EEG

Cognitive Research

  • Experiments with long- duration stimuli (i.e. task requiring sustained attention, ecologically appropriate stimuli)

  • Monitoring sleep stages

Clinical Research

• Epilepsy (Detection of seizures, Localization of focus/foci, Prediction of seizure onset)

• Monitoring the level of anaesthesia

• Detection of brain death

• Measurement of drug effects

• Detection of cerebral pathology, e.g., through blood supply problems

• Sleep disorders

• Almost all neurological disorders have EEG correlates 

 

18 of 33

Event Related Potential (ERP) and Evoked Potential

General class of potentials displaying stable time relationship to a definable reference event

  • Reference event  (Onset/offset of a stimulus, Motor response, Decision moment)

  • Terminology

    • –  EP: Perception and clinical research

    • –  ERP: Experimental cognitive research

  • ERPs are waveform characterized by a series of positive (P) or negative (N) deflections at different latencies ERP Components

Exogenous Components: Modulated by external characteristics of stimuli Endogenous Components: Modulated by internal characteristics

(But this distinction is not definite) 

Advantages:

  • ERPs are simple, fast to compute

  • ERPs require very few analysis or parameters

  • ERP has high temporal precision and accuracy

  • ERP literature is quite mature

  • ERP provides a good quality check 

19 of 33

ERP Model: Signal Averaging

ERP Hypothesis: ERP is a signal (s) that appears superimposed and without interaction on the background or ongoing EEG which is considered random noise (n).

Assumptions:

• ERP is uncorrelated with background EEG
• Background EEG is random
• ERP is invariant across trials (same ERP is repeated over trials)

• Background EEG varies (randomly) from trial to trial

After averaging across trials, noise will cancel out and only the event related EEG response will remain. 

How many trials?

As many as possible

The number of trial depends on signal-to-noise characteristics

the effect size
the type of analysis to be performed

R = amount of noise on a single trial N = total number of trial

SNR (signal-to-noise-ratio) increases as a function of the square root of the number of trials 

20 of 33

ERP: How many trials?

Practical suggestions

– 50 trials / condition / participant

– Similar number of trials for all conditions (Phase/power produce positive bias with fewer trials)

– If not possible, match trial count

  • Select the first N trials from each condition (N = the number of trials in the smallest condition)

  • Select N trials at random

  • Select N trials based on some relevant behavioural or experiment variable (i.e. reaction time) 

Signal to noise calculations/rationale:

For example, you are interested in P3 component and its amplitude is 20 μV

Say, noise in a single trial EEG is 50 μV

The S/N ratio on a single trial = 20/50 = 0.4

The S/N ratio of two trials = (2) x 0.4 = 0.56

The S/N ratio of 20 trials = (20) x 0.4 = 1.79

The S/N ratio of 200 trials = (200) x 0.4 = 5.66

Therefore achieving a substantial increase in S/N ratio requires a very large increase in the number of trials.

So improve the quality of the data by decreasing the source of noise than by increasing the number of trials 

21 of 33

ERP components

An ERP component is a part of waveform with a circumscribed scalp distribution (physiological substrate) and a circumscribed relationship to experimental variables (functional substrate).

Examples:
MMN (mismatched negativity, 160-220 ms at central sites) N170 (face-related potential at occipital sites)

Why study ERP components?

1. Common language linking diverse experiments, paradigms etc

2. Base for integrating ERP with other measures of brain activity

3. Structure-function information 

“Baseline” period: In averaging, all trials are set (arithmetically) to have the same zero voltage at stimulus onset, so that only deviations from the baseline voltage are seen in the ERP, after stimulus presentation.

Baseline subtraction (mean of baseline period is subtracted) 

22 of 33

Limitations of ERPs

Limitations of ERPs

There are two main limitations.

  • The first concerns interpretational issues, particularly with regard to interpreting null results

    – ERP reveal little of EEG information
    – ERP does not capture non-phase-locked responses

  • The ERPs provide limited opportunities for linking results to actual neurophysiological dynamics

    – ERPs are less understood compared to the neurophysiological mechanisms that produce neuronal oscillations and synchrony 

23 of 33

Applications: Visual Evoked Potentials (VEPs)

Presenting stimuli at intervals longer than a sec elicits a transient VEP

Presenting stimuli at faster rate elicit Steady State VEP (SSVEP)

• Provides fine grained information of early visual processing

• Routinely used to assess the integrity of the visual system from retina, via optic chiasm, to primary visual cortex

• Used in several clinical cases (Multiple sclerosis, Visuospatial neglect patients, Localization of optic lesions)

24 of 33

Applications: Auditory Evoked Potentials

Early AEP/ Brainstem AEP

• Testing the integrity of primary auditory pathways • Localizations of deficits/damages
• Detection of early hearing loss in infants

Middle Latency Potentials

10 ms < Latency < 50 ms
• Possibly thalamic and very early cortical responses • Earliest attention related effect (P1)

• Monitoring purpose in anesthetized conditions 

Late Latency Potentials

• Cortical origin
• Reflect both the sensory processing and ongoing cognitive processing (i.e. attention)
• N1-P2 complex
• Developmental language disorders, dyslexia 

25 of 33

Somatosensory Evoked Potentials (SEP)

Various Sub-components

Early N10 reflect action potentials from the peripheral nerves
• Subcortical (thalamic) components
• Cortical components for later latencies

Applications

•Tracking the somatosensory pathways from periphery to cortex

• Estimation of peripheral conduction velocities

• Patients with spinal chord injury, traumatic lesions, multiple sclerosis

• Classification of patients (i.e. apallic syndrome from locked-in states)

• Neuromonitoring purpose (spinal chord) during surgery 

26 of 33

Chemosensory Evoked Potentials

• Issue of timing

• No clear presence of early ERP components

• Amplitude is modulated by the concentration of odorant

• Useful for ageing study (olfactory discrimination degrades with normal ageing)

• Early detection of Alzheimers (olfactory impairment precedes cognitive impairment in AD) 

27 of 33

Contingency Negative Variation (CNV)

  • Indicator of learning paired stimuli (Get Set – Go)

• Reflection of attention, concentration & readiness to S2

• Index of neuronal excitability 

28 of 33

Music

Music: Rule Violation

  • Music, like language, has “rules” – called syntax

  • A listener expects specific musical events according to a preceding musical context

  • Ex: Chord functions with final chord as regular (harmonically expected) or irregular (unexpected)

    Sample-1 Sample-2

    Remember: The last chord itself is a valid musical entity, only the prior context makes the difference! 

.................

Similar semantic priming effect in language and music

Medial temporal gyrus (neighbor to STS) - Center for semantic integration 

29 of 33

Time Frequency Representations

Advantages:

  • Clear interpretations
    – Neurophysiological mechanisms – Ubiquitous oscillations

  • Neuronal oscillations are the most promising bridge linking findings from multiple disciplines

  • Covers a more comprehensive multidimensional space 

Disadvantages:

  • Decreased temporal precision

  • Complicated analysis strategies

  • Fewer previous research for contextualization of findings

  • Does not provide information on the co- operation between brain regions 

30 of 33

Neuronal Synchrony

  • Cognition requires cooperation between neural populations within and across brain regions

  • Synchronization of neural oscillations as a mechanism for integration of neural populations mediating perceptual binding and cognitive brain networks

    – Neuronal assemblies that oscillate in synchrony exchange information more effectively relative to nonsynchronously oscillating assemblies

  • Synchronization between multiple and distant brain regions

Measures of Synchronization Nonlinear methods:

Nonlinear methods:

  • Nonlinear correlation
  • Information Theory – Mutual Information – Transfer entropy

  • Phase Synchrony - Hilbert +Shannon – Mean phase coherence – Wavelet

  • Generalized Synchrony – Similarity index & families – Mixed predictability
    – Cross prediction

Linear Methods:  

  • Linear Correlation

• Coherence – Magnitude squared coherence – Partial coherence

  • Granger causality

  • Multivariate modeling – Directed transfer function – Partial directed coherence 

 

31 of 33

EEG Issues

Experimental

  • Choice of reference

  • Volume conduction

  • Relation between ERPs and Ongoing EEG oscillation

  • Trial by trial variability

  • Effect of pre-stimulus brain responses

    Analysis

  • Single trial analysis

  • Synchrony in the source space 

32 of 33

.

33 of 33

Comments

No comments have yet been made

Similar Psychology resources:

See all Psychology resources »See all Lecture notes by JB resources »