EEG summary Flashcards
Event-Related Potentials (ERPs)
ERPs are voltage changes in the EEG that are time-locked to specific events or stimuli. They represent the brain’s averaged electrical response to a particular type of stimulus.
Key steps in ERP analysis
A. Preprocessing
B. ERP formation
C. ERP analysis methods
Preprocessing
- Re-referencing
- Filtering
- Artifact Handling
Re-referencing
A technique in EEG analysis that establishes a neutral reference point for voltage measurements. Common methods include mastoid reference, average reference, and nose reference.
Re-referencing impact
Choice of reference can affect ERP morphology but not topography
Filtering
The process of removing unwanted frequencies from the EEG signal. High-pass filters eliminate slow drifts, low-pass filters remove muscle artifacts, and notch filters suppress power line noise (50/60 Hz).
Filtering - important considerations
- Filter settings can distort ERP components
- Different filter settings can produce different ERP morphologies
- Always report filter settings in publications
Artifact Handling
- Ocular artifact correction
- Artifact rejection
Ocular Artifact Correction
ICA (Independent Component Analysis)
- Separates EEG into independent components
Identifies and removes eye movement components
- Reconstructs clean EEG
Regression-based methods
Artifact rejection
- Automatic: Based on amplitude thresholds
- Manual: Visual inspection
- Criteria should be consistent across conditions
ERP formation
- Segmentation
- Baseline correction
- Averaging
Segmentation
The process of cutting continuous EEG data into time-locked epochs around a stimulus, such as -100 to +1000 ms. Each condition should have unique trigger codes
Baseline correction
A technique that subtracts the mean pre-stimulus activity (e.g., -100 to 0 ms) to normalize EEG epochs. Ensures all epochs start at approximately 0 μV
Averaging
- Average all epochs per condition
- Improves signal-to-noise ratio
- Number of trials needed depends on component size and noise level
ERP Analysis Methods
- Peak analysis
- Mean amplitude analysis
Peak analysis
A method that measures the maximum or minimum voltage within a time window to analyze ERP components.
Peak analysis - advantages
- Simple to implement
- Widely used and understood
Peak analysis - disadvantages
- Sensitive to noise
- May miss sustained effects
- Assumes clear peak exists
Mean Amplitude analysis
A method that calculates the average voltage within a time window. It is more robust to noise than peak analysis.
Mean amplitude analysis - advantages
- More robust to noise
- Better for sustained effects
- No assumption about peak presence
Mean amplitude analysis - disadvantages
- May miss brief effects
- Window selection can be arbitrary
Frequency analysis
A. Fourier transform methods
B. Common frequency bands
C. Specialized Measures
Basic fourier transform
- Decomposes signal into constituent frequencies
- Shows power at different frequencies
- Loses temporal information
Time-frequency analysis
Methods like wavelet analysis and Short-Time Fourier Transform (STFT) that provide both frequency and temporal information about EEG signals. Shows how frequency content changes over time
Common frequency bands
EEG signals are categorized into frequency bands such as
- Delta (0.5-4 Hz): Sleep, deep relaxation
- Theta (4-8 Hz): Memory, emotional processing
- Alpha (8-13 Hz): Relaxed wakefulness, inhibition
- Beta (13-30 Hz): Active thinking, focus
- Gamma (>30 Hz): Feature binding, high-level processing
C. Specialized measures
- Event-Related Synchronization/ Desynchronization (ERS/ERD)
- Frontal Alpha Asymmetry (FAA)
Event-Related Synchronization/ Desynchronization (ERS/ERD)
Measures changes in power relative to baseline, used for studying cognitive and motor processes.
Frontal Alpha Asymmetry (FAA)
A measure comparing alpha power between left and right frontal regions, used in emotion, depression, and personality research.
Multivariate Pattern Analysis (MVPA) - Basic principles
- Uses machine learning to identify patterns in neural activity
- Considers multiple channels/timepoints simultaneously
- Can detect subtle differences between conditions
Multivariate Pattern Analysis (MVPA) - Common approaches
- Classification
- Train classifier to distinguish between conditions
- Cross-validation to test generalization
- Measure classification accuracy - Representational Similarity Analysis (RSA)
- Compare patterns of neural activity
- Create similarity matrices
- Link neural patterns to behavioral or theoretical models
Multivariate Pattern Analysis (MVPA) - Advantages
- Can detect distributed patterns
- More sensitive than univariate methods
- Links directly to information content
Advanced analysis methods
A. Source localization
B. Connectivity analysis
Source localization
Techniques like LORETA and BESA estimate the neural sources of scalp EEG activity but require accurate head models due to the inverse problem.
Connectivity analysis
- Coherence
- Measures frequency-specific synchronization
- Shows functional connectivity between regions - Phase Synchrony
- Measures phase relationships between signals
- Independent of amplitude - Granger Causality
- Assesses directional influences between signals
- Based on predictability
Best practices for method selection
- Research question considerations
- Data Quality Requirements
- Validation strategies
- Reporting standards
Research question considerations
- Temporal precision needed
- Spatial information required
- Type of neural process (evoked vs. induced)
- Expected effect size
Data quality requirements
- Number of trials needed
- Signal-to-noise ratio
- Artifact levels
- Recording parameters
Validation strategies
- Multiple analysis approaches
- Control analyses
- Statistical validation
- Cross-validation where appropriate
Reporting standards
- Detailed methods description
- Parameter specifications
- Preprocessing steps
- Statistical approaches
