EEG: BCIs

Question 1

Electrical brain activity in BCIs is acquired in 3 different ways:

Answer 1

○ EEG (electroencephalography)
○ ECoG (electrocorticography)
○ Intracranial microarrays

Question 2

Two main types of BCIs, based on:

Answer 2

1. Spontaneous oscillations (e.g., the visual alpha rhythm - eyes closed - active)
2. Event-related potentials (ERPs; e.g., attention-modulated P300)

Question 3

What does it mean that most rhythms are idle rhythms? Examples.

Answer 3

They are attenuated during activation. Examples:
○ Alpha-rhythm (around 10 Hz) in the visual cortex - activates when eyes are closed and chills when they’re open
○ Sensory-motor rhythm (SMR; around 14 Hz) in motor and sensory cortex - activates when arm is at rest, chills when arm moves

Question 4

BCIs based on ERPs examples (4)

Answer 4

Oddball paradigm, The Matrix Speller, Hex-o-Spell, Spelling with Spatial Auditory Attention

Question 5

Oddball paradigm:

Answer 5

• A sequence of symbols is presented with the task to attend the occurrence of a target stimulus, while ignoring the others
• The EEG shows a specific pattern (P300) after attended targets
• The trick: This allows the BCI to conclude which symbol the user wants to select.

Question 6

The Matrix Speller:

Answer 6

• Selecting letters in one step
• The letters (or rows and columns) are flashed in random sequence.
• The user attends, e.g., the letter B
• The flashing of the target letter elicits the P300 which can be recognized by the BCI. Other letters elicits a much weaker P300
• Gaze control necessary!

Question 7

Hex-o-Spell:

Answer 7

• Selecting letters in two steps
• Letters are placed into groups
• The user first selects the group, then specific the letter

Question 8

Spelling with Spatial Auditory Attention:

Answer 8

• Similar principle to the Hex-o-Spell
• User learns which tones are linked to which letters
• BCI speller is based on spatial hearing, and no visual processing

Question 9

Challenges in EEG (4):

Answer 9

Volume conduction in EEG, Subject-to-subject variability, Session-to-session variability, Trial-to-trial variability

Question 10

Explain the Volume conduction in EEG challenge

Answer 10

• The signal arrives with almost equal intensity at different scalp locations due to the different tissue conductivities
• Spatial smearing: raw EEG scalp potentials are known to be associated with a large spatial scale owing to volume conduction
• In this typical example data set, most of the channels are highly correlated
• The map shows the correlation 0.5 coefficient of each channel with channel Cz in the center

Question 11

Explain the Subject-to-subject variability problem

Answer 11

• The human brain is very plastic, in particular during early development (e.g., hemispherectomy during childhood)
• BCI illiteracy:
  □ Large dataset of over 150 participants using a SMR-based BCI
  □ An estimated 20 - 30% of the population are not able to modulate their SMR
  □ Reasons remain unclear, but potentially due to tissue conductivity differences

Question 12

Independent Component Analysis (ICA):

Answer 12

Extracts components that are maximally independent to find likely sources in the brain (or be identified as artifacts)

Question 13

Linear Discriminant Analysis (LDA)

Answer 13

○ Well known classifier used to separate two or more classes of subjects
○ Uses means of the two classes and the noise (covariance of two classes), to find the separating line (between the two classes), the direction of noise and separate two classes
○ Given two Gaussian distributions 𝒩 (𝜇1 , Σ) and 𝒩 (𝜇2 , Σ), LDA is defined by the normal vector

Question 14

Can a spontaneous voluntary movement be cancelled after the onset of the readiness potential?

Answer 14

Spontaneous, voluntary movements can be cancelled after onset of the readiness potential, however only until a point of no return at around 200 ms before movement onset