EEG General

Question 1

What is data pre-processing and what are the steps?

Answer 1

Pre-processing is a crucial part of the analysis and consists of multiple steps

1. data reorganisation (epoching)
2. removing bad channels (depends on the design, protocol, EEG system, planned analysis and the protocol)
3. Keeping a record of everything removed and raw data
4. The same steps are repeated in the same order for all participants

Question 2

What does pre-processing involves?

Answer 2

• import the data
• inspect the data
• rereferencing
• channel interpolation
• high pass filtering
• extracting data epochs
• artefact rejection
• artefact corrections

Question 3

What is voltage?

Answer 3

It is the difference of electrical potentials between two electrodes

Question 4

What is re-referencing?

Answer 4

It is done when the electrodes used in the recording is not the site of reference that is desired in the analysis.
It is done by substracting the average activity at the reference electrodes from all channels

Question 5

What is Channel interpolation?

Answer 5

The data from missing electrodes is estimated based on the activity and location from other electrodes. E.g. nearest neighbour, linear, spline interpolations

Question 6

Why is channel interpolation done?

Answer 6

Preferred to removing bad channels as it keeps the data matrix the same across participants
It is necessary for surface Laplacian
Does not create new data, therefore reduces the rank of the data matrix

Question 7

What is high-pass filtering?

Answer 7

High frequencies pass the filter above a certain frequency while low frequencies are filtered out as they are usually noise
applied to continuous data to avoid filtering artefacts at the edges of the data segments
lower edge is set to 1Hz to get rid of activity with a one second or longer cycle

Question 8

What is epoching and when is it done?

Answer 8

It is done when investigating task-related changes
EEG data is continuous and recorded as a 2D ,matrix (time x electrode) and cut into segments surrounding events
after epoching the data is 3D matrix (time x electrode x trial)
Not needed for resting state analysis at the preprocessing stage
Need to consider what are the events of interest, latency of the components of interest, the baseline period

Question 9

What are some types of artefacts?

Answer 9

eye-blinks - low frequency and large amplitude, can be
correct problematic for analysis
saccades - frequency profile may overlap with the EEG components of interest
muscle activity - electrodes at the edge of the scalp, high frequency, can use low-pass filter to filter them out
Strong alpha activity (8-12Hz) - indicate tiredness, cannot be filtered problematic for ERP analysis and potentially for TF if one condition is more arousing than the rest
line noise from oscillating voltages - extrernal sources, can use notch filter (band filter) or low pass filter. less problematic for ERP, more problematic for TF (even though noise can be assumed as a constant)

Question 10

What is artefact rejection?

Answer 10

detect artefacts through visual inspection or semi-automatically based on a criteria such as

• amplitude threshold (-75mV)
• min-max threshold
• low activity
• spectral distribution
• standard deviation
• joint probability

Question 11

What are some problems associated with artefact rejection?

Answer 11

Loss of a large quantity of the data
Need to truly reject only channels that carry artefact data
some individuals may be more prone to certain types of artefacts
can use independent component analysis , which is a class of blind source separation techniques

Question 12

What is independent component analysis (ICA)?

Answer 12

A class of blind source separation techniques
It is independent because the sources of the signal are unrelated, e.g. one signal is not related to the other and the information about the value of one signal does not carry information about the value of the other