Post on 14-Apr-2017
Biophysics of EEG
Outer Surface of Human Brain
Gray Matter = Neuron cell bodies & dendrites
White Matter = Myelin (=fat)- covered axons
Cortex = Outer layer of gray matter
Gyri
Sulci
Axial Slice through Human Brain Front
Back
Closed and Open
Electrical Fields
Single pyramidal cell in cortex
For brain electrical activity to be detectable through skull, must be strong signal summed over many neurons - All behaving similarly at same time - All oriented in same way - So negative and positive don’t cancel each other out when summed Pyramidal Cells in the cortex have the right properties
Pyramidal Cell Orientation in Cortex
Pyramidal Cells mostly oriented in same direction relative to cortical surface, pointing out toward surface - Electrical activity from gyri more detectable at scalp than activity from sulci
Tangen'al Dipole
Radial Dipole
Magne'c Field
Electrical Field
Electrical and Magne.c Fields Generated by Dipoles
DIPOLE
Recording Brainwaves = Electroencephalography (EEG)
Digitize = Take samples of continuous input at regular intervals (e.g., every 5 msec = 200 Hz)
= Timing of critical “events” - Used in extracting Event-Related Potentials (ERPs) from ongoing EEG
Easy Cap
Alpha Waves
EEG Rhythms
Left front to back Right front to back
Left front to back
Right front to back Midline front to back
Left front to back Right front to back
Left front to back
Right front to back Midline front to back
“Inverse Problem” in
EEG Source Localization • Electrical fields are “volume-conducted” throughout
head o Conducted better through some kinds of tissue
• E.g., Cerebrospinal fluid vs bone
• And summate o Both boosting and canceling
• So cannot assume activity generated directly under where it’s detected on scalp
• Patterns detected at scalp could be produced by many different combos of generators o So insufficient by itself to reveal locations of sources
Event-‐‑Related Activity
Recording “epochs” Scale & Polarity (Negative Up!!!)
- Input range = 40-50 microvolts (uV) - Activity of interest often only 1-5 uV, so buried in higher amplitude activity that’s not of interest
(We record continuously & pull epochs out later)
Averaging Event-‐‑Related Potentials (ERPs)
“Oddball” Paradigm = 1 stimulus appears more often (X) than than another (O)
Typically see large “P3 (P300)” in response to “oddball” stimulus
P300 can sometimes be big enough to see in individual trials - Not true for N400
Alpha Waves (8-12 Hz)
Idealized Auditory ERPs
Electrical potential from a single source in left thalamus volume-conducted through white matter
Some ERP Peaks Related to Language
N400 = ERP “component” related to meaning - Bigger when word’s meaning doesn’t fit context - Bigger for unfamiliar words - May reflect amount of work required to integrate with context
P600 = ERP “component” related to form - Bigger when word not of expected type for a position in a sentence - May be a type of P300 - Sometimes called Syntactic Positive Shift (SPS)
Can evoke both at once
Alternative Signal Representations and Transformation
• Signal representation in a suitable domain for information extraction,….
Examples: o Fourier Transform for spectral analysis o Hilbert Transform in envelop detection o KLT(PCA) for optimal function approx. o Laguerre basis function o Numerous other transforms( DCT,Radon,…)
Two Domains for Signal Representation
Different projection spaces different signal representa-tions
Signal Domain
Projection space, Basis functions
Transformed Domain Transformation
Main Stages in Signal Analysis Signal domain basis functions Transformed domain, Coeffs
Transforma)on
Analysis
Reconstruc)on
Signals
Recon Signal
basis functions
Modified Coefficients
Information
extraction
Representation in Time f(t)
Representation in Freq F(ω)
Time Domain Studies Frequency Domain Studies
Fourier Transform