compared to reference electrode

--> mastoid bone, vertex etc
--> limitation = if recorded close to that location eg mastoid bone, the deflections will be smaller because the reference electrode will pick up part of the same activity !! #

noise reduced by electrode at eye muscle

--> eye blinks usually interfere with alpha activity and have 7.5Hz

detecting abnormalities

--> Well established know pattern compared against measurement
--> any deviation might imply differences in processing

Analyzing timing of cognitive processing

--> low spatial resolution prevents localization
--> done through averaging EEG waves before stimulus is presented and comparing it to average of EEG waves after stimulus is presented (ERP) #

Gamma waves (36 to 44Hz)

--> usually 30hz to 80hz, but
more difficult to measure cause skull mutes the high frequency signal + lots of noise
--> directly associated with increase in brain activation
--> object recognition, top-down processing of sensory input, perceptual binding (integrating various aspects of stimulus into one)

averaged because:

--> ERP = tiny signal (2 micro volts while EEG 20micro volts)
--> eliminates noise (brain activity not related to stimulus + measurement errors)
--> by averaging all the other activity out :3!!
--> leads to constant signal

Positive deflection (P)

--> positive rising wave is deflected at the peak into negative

Sometimes they are just labeled according to order of occurrence instead of time

--> eg. N1 happens first, then P2, then P3

peak starting 100 ms

--> sensory processing modulated by attention
--> N100 + P100 = selective attention (bigger amplitude when attending vs not attending)<-- visual attention task :3 !!(attend left ignore right and vice versa)
------> Supports early selection model of attention :3 !!

peak starting 300ms

--> P300 = stimulus present that was paid attention to (especially if rare) #

By looking for timing differences in processing of stimuli some illnesses can be deduced (eg sclerosis, tumors in auditory system)

Limitations

--> speficic localization near impossible cause only signal from scalp and bad spatial resolution

tangential

--> from sulci
--> parallel/horizontal to the inside of the skull (like lying flat on their side and are paralallel to the inside of the skull like a blanket :3 !!
--> cant be measured by electrodes directly above (cause positive negative extracellular fluid cancel each other out)
---> thus can only be measured by electrodes to the side of it :3 !!

radial

--> from gyri
--> radiating from center of brain outwards towards the skull
--> perpendicular to the curved shape of the skull (like standing on it upright!)

Which then leads to a positive deflection on the EEG

which leads to a negative deflection

--> if not parallel but random order = no clear dipole possible = no signal

--> if not oriented the same way = positive and negative extracellular fluids mix and cancel each other out = no signal

Dipoles must be active at the same time (Synchronous)

--> otherwise the signal will be too weak for the electrodes to pick it up
--> mediated by thalamocortical / corticocortical connections

measurement (positive / negative deflection) represents the amount of ions (positive negative) at the electrode at that time pushed or pulled there by the dipoles
--> Dipoles exude influence in all direction

How quickly the amount of electrodes switch from positive to negative at the electrode, determines the frequency
--> usually 1 to 30 Hz (Gamme 30 to 80 hz)

Spatial smeering!

Dipoles exude influence (repell / attract in all direction)

--> 1 dipole will affect multiple areas of the scalp!
--> this is called Spatial smeering!
--> happens for main eeg range 0.1 to 30hz, for higher frequency it doesnt happen

the stronger the dipole the further they can "send" information
--> possible electrodes could pick up deeper cortical tissue activation (theoretically as dipole output weakens exponentially over distance)

Enhancing measurements

--> by shielding the electrodes / cables from outer influences
--> by filtering out muscle movements from eye eg. or breathing contraction

--> both reduce noise :3!!

50 to 150ms long latency repsonse
--> p50(P1), n100(N1), p160(P2)

N170 (Part of N100/N1 component)--> bigger for faces than not faces --> bigger for birds than non birds in bird experts etc..

Alpha waves (8 to 13Hz)

--> predominant during resting (especially with eyes closed)
------> first eye closed, then sudden eye opening = disrupts alpha waves (alpha desynchronization)
--> greatest above posterior occipital/temporal regions and parietal regions
--> associated with information processing
--> 8 to 10Hz = associated with stimulus-unspecific and task-unspecific increases in attenttion

Caused by interaction of thalamus and cortex

--> e.g thalamus activation/osszillation of 7.5 to 12.5Hz activate osszillation in cortex <-- OHH WOW REALLY lol thalamus = heavily interconnected lawl noobs!!

Widepsread scalp distribution

--> drowsiness
--> less information processing (lol cause drowsieness lool xD sleepy kitty ahw !!)

Spectral analyses provides important information about the frequency compositions of EEG oscillations.

--> BUT!!! does not provide temporal information as to when frequencies occur

assumptions of stationary signal

--> thus only 3.5second long snippets , but 60 seconds in total to reduce second to second variability in the signal !!

absolute vs relative power

--> absolute = amount of frequency in whole EEG
--> relative = amount of activity in frequency divided / by the the total power of the frequency wave

10/10 system
--> also includes electrodes betwwen the position of the 10 /20 system

5/5 system
--> even more electrodes between the 10/10 system
--> diminished return in accuracy increase after 100 electrodes

even distribution of electrodes = very important regardless of system!!

there is no perfect reference electrode

-->at skull always close to other electrodes,
-->below skull too much interference from muscles (especially heart muscle)

Limitation

--> BUT!!! does not provide temporal information as to when frequencies occur

sampling (Nyquist Theorem)

As a general rule, the sampling rate should be at least twice the highest fre- quency present in the signal under investigation. This rule, also known as the Nyquist Theorem,

linear (nearest neighbor)

--> uses a weighted average of the nearest electrode to make up for missing electrode value

spline

--> Uses the average of all electrodes on scalp to make up the missing electrode value #

average reference approach

--> averaging activity by all electrodes and use that to compare activity if each electrode agains <-- average reference approach

hjorth method / source derivation

--> averaging the 4 nearest electrodes and compare the single electrode to that average, this happens for all electrodes (so their nearest 4 average is what their activity is begin compared against :3) --> hjorth method / source derivation #

Sphere model

if head was a sphere, and all tissue/ bones equally conductive it would be easy to know how the voltage from a signle dipole would distribute at the scalp (thus easier to reverse engineer back to source (dipole))

----> BUT' head is not a sphere and tissues and bone structure conduct electricity differently

Assumptions

--> resistance / conductivity in each voxel is assumed to be same (ok since they are super small)
--> resistance/ conductivity between the voxels differs!!

resistance / conductance is calculated for each individual voxel

--> based on resistance values obtained from dead tissue or animal tissue

Now follows real shape of head and real resistance values

--> distribution of electric activity of dipole on scalp can be easily calculated :3 !!

Assumptions!!:

--> most boundaries have different resistance /conductivity (brain, dura, skull, skin)
--> within the boundary the resistance/conductivity = same

resistance / conductance is calculated for each individual boundary

--> based on resistance values obtained from dead tissue or animal tissue

General approach

--> brain divided in 100 voxels
--> each voxel = 3 dipoles (up, forward, lateral)
--> strenght of the voxels (thus dipoles) adjusted until voltage distribution matches the measured one

--> limitation:

• 100*3dipoles= 300 dipoles = 300 electrodes needed
  
  • high processing power
• dipoles not independent of each other

cortcally constraint

--> brain surface = divided into tiny triangles with 1 dipole each (perpendicular to cortical surface)
--> strength of triangles (dipoles) adjusted until voltage distribution matches measured one

--> limitations:
--> in sulci, dipoles can become parallel to each other, thus if oriented oposingly could cancel each other out or magnify each other which distorts model
--> non uniqueness!!

Minimum norm solution

--> builds on cortical constraint
--> says model that best conforms to measured one, but with least amount of high peaks = best one (so basically ignores some activity in sulk where the spikes occur tbecause of being parallel to each other thus stronger signal!)
---> ACTUALLY always finds a unique solution to inverse problem

--> limitations:
---> biased toward dipoles closest to scalp
-------> can be corrected with depth weigthed minimum norm # #

assumptions:

--> voltage/ current distribution can be modeld by less than 20 dipoles
--> each dipole = fixed but strength varies over time ?? <-- nope cause orientation = also adjusted

arbitrarily choosing whatever amount of dipoles researcher pleases and putting them in whatever position and location the researcher pleases , strength of dipoles also randomly chosen

--> also two biggest limitations here lol

calculates how well the model compares to actual measured voltage distribution

--> adjust orientation and strength slightly to aproximate it more
--> test again
--> adjusts again
--> etc until good fit found (UNIQUENESS PROBLEM!!)

-------> study where EEG voltage (by 10 dipoles unknown to participants) given to participant researchers, each supposed to find out how many and where localized.
--------------> semi accurate placement within 1 to 2 cm, but either too many or way too little actual dipoles set 6 to 12 lol)
--> shows no unique / validifyable way of localizing dipoles!

low resolution electromagnetic tomography (LORETA)

--> says that voltage emitted by dipoles decreases gradually as it travels through the brain and towards scalp
--> thus selects the voltage distribution of high activities that is as smooth as possible.

---> limitation:
--> boundaries between areas blurred, <-- thus only usable for finding center of activation, where ERP originated

P3a

--> occurs for unexpected stimuli (like in the oddball task)
----> e.g. count all the X's (80% X's) But also 20%O's then P3a activity super high for the unexpected O's :3!! (also basically thats the oddball task :3 !!
--> frontal scalp distribution

--> in schizophrenia P300 delayed by 75ms and 50% weaker than healthy controls (slower response selection!!)

Endogenous cuing

--> attention task = fixate on cross, arrow will indicate which side stimulus will be
---> RT faster if stimulus will be in indicated spot, cue gave us the idea where it will be so we could consciously ready ourselves to react to it (Conscious processing = endogenous )!! #

**--> affects P100/P1 and N100/N1 (amplitude bigger if attended location is wehre stimulus will be )

Exogenous cuing

--> attention task = fixate on cross, arrow will indicate side where stimulus will be
--> Now instead of arrow, a lightning flash will be displayed randomly in either visual field. --> this grabs their attention toward that location for 50 to 200ms and makes reaction time to stimulus faster , if it appears i that location within that timeframe (<-- because flash random and gives us no cue where stimulus will be it is exogenous, we don't have time to think about :)!)

---> if stimulus shown after 50 to 200 ms, the reaction time to stimulus if shown in same area as flash is Inhibited / slower (inhibition of return) #

--> affects P100/P1 component