Source Localization

How many electrodes are needed for source localization?

at least 64, ideally 96

spatial nyquist

rule describing how the more sources there are, the more electrodes you need to pick them up

Why is the shape of EEG/MEG waves on the scalp important?

it can indicate the number of sources, or at least help discern it

forward problem of source analysis

creating a scalp topo map of known neural sources, usually quite doable, only a limited number of solutions

inverse problem of source analysis

the issue that arises mathematically when you want to find the neural sources for a particular topo map of brain activity. There are an infinite number of solutions each time because you don't know how many dipoles there are

What characterizes a current dipole?

position, orientation and strength. Orientation is especially important

What is the general analysis strategy of dipole source analysis?

1. use scalp maps to study the spatio-temporal characteristics of ERP
2. make a model of the generators (current dipole model, choose a number of dipoles)
3. compute EEG/MEG then evaluate fit and change parameters (forward solution, how good is the model for your data)
4. think critically about whether these sources actually could represent the data (inverse problem)
- most of this is done by computer programs

What kinds of head models exist for source analysis and scalp fitting?

1. spherical head model (or could use ellipse)
2. standard head model
3. realistic head model for the participant

How many layers do head models have? Why?

three, brain, skull and scalp, since these different tissues have different conductance and thus change the electrical output

What are the benefits and drawbacks of spherical head models?

- they are cheap and easy to use
- sometimes they don't fit the head well since the head is not a sphere

What are the benefits and drawbacks of standard head models?

- easy to handle
- more realistic than spherical models since they are based on real heads
- makes the results comparable between subjects
- relatively cheap
- lose some accuracy due to averaging, resolution of brain is very fuzzy

What are the benefits and drawbacks of individual head models?

- very high accuracy (based on real subject MRI)
- super expensive
- low comparability between subjects

What have simulation studies of ERP source localization shown in regards to accuracy?

- with a 10 dipole source (this is massive, many wouldn't even localize on this many)
- the average error was 1.4cm for source, very good, so when dipoles were localized, they were localized well
- but models often combined dipoles that were 5cm apart, so the biggest issue was underestimating the number of dipoles

How can source localization be improved?

- make physiological hypotheses
- use PCA to combine electrodes to select major factors to move forward with
- fit prominent peaks in the recording
- go off previous research
- use neuroimaging results

How many dipoles are often considered the maximum to do source analysis on?

3

regional source

a neural generator of activity that includes 3 dipoles in orthogonal directions to account for all orientations of neurons in a fold of cortex. Taking this into account often increases model fit for source analysis

minimum norm approach

method used in distributed source analysis, states that the solution has the lowest overall source activity (is most realistic to brain activity). Biased towards superficial solutions, but can be corrected with depth weighting

LORETA approach

assumes that source voltages change gradually across the brain, so neighboring neurons should have similar activities. However, it ignores sharp changes between neighboring functional regions resulting in over-smoothed images

induced response

a change in neural oscillations that is time-locked but NOT phase locked to an event, not seen in an averaged ERP

What methods can be used to study induced responses?

1. time frequency plots
2. temporal spectral evolution
3. event related synchronization or desynchronization
4. oscillatory coupling/coherence