6. MEG Flashcards
What does SQUID stand for?
Superconducting Quantum Interference Device
Electrical current flow in a wire produces a magnetic field _____ to the current flow…
perpendicular
- Think of the right-hand rule where thumb points in the direction of the current flow and the fingers wrapping around the wire determine the direction of the magnetic field
Which neurons does MEG focus on?
Pyramidal neurons (they are arranged in columns in the neocortex)
What are radial fields and what is one issue with detecting them?
they are magnetic signals produced by neuronal activity at the top of the gyri
- they are less detectable by MEG
What are tangential fields?
they are magnetic signals produced by neuronal activity in the sulci
- they are detectable by MEG
What are the potentials we primarily measure with MEG and how do they differ to the ones we measure with EEG?
Post-synaptic potentials (these can be excitatory or inhibitory) DON’T CONFUSE WITH ACTION POTENTIALS
- When an impulse arrives at a synapse from an activated presynaptic neuron, a neurotransmitter is released causing channels in the membrane of the postsynaptic neuron to open creating a shift in the resting membrane polarization.
This polarization leads to activation or inhibition of voltage-gated channels responsible which cause action potentials
In MEG we measure the strength of the ______ (___)
magnetic field (in Tesla)
- aka magnetic flux density
What is magnetic permeability (μ0)?
How much a material magnetizes if we apply a magnetic field to it
Ferromagnetic materials
Retain magnetism after being magnetised (really bad around fMRI scanners)
Paramagnetic materials
Amplify magnetic fields but do not stay magnetized (make magnetic fields bigger when passing through them)
Diamagnetic materials
Weaken applied fields (repulse them, will do anything to drive a magnetic field to 0)
How is MEG cooled?
liquid helium
What are superconductors?
Materials which lose all of their electrical resistance when cooled to a low enough temperature
- So, if we put a current into a superconducting loop, it will continue infinitely (as long as we maintain the low temperature)
- In reverse, applying a magnetic field to a superconducting loop will induce a current to cancel out the field
SQUIDS in MEG use such superconductors (~4K/ -269.15C)
Magnetometers…
measure the magnetic field (a coil with wrapped one way in a circle - 1 loop)
- very prone to background noise
- more sensitive to deeper parts of the brain
Gradiometers…
Measure the magnetic field (a coil wrapped into two loops going in opposite directions)
- more of a noise resistant measurement
- comes at a cost of being less sensitive to deeper parts of the brain
How does MEG work?
Magnetic signal present during post-synaptic potentials induces a current in the pickup coil.
This current is transferred to another coil located just under the squids where it is turned back into a magnetic field which can be measured by the SQUID
- this mechanism is known as a flux transformer
A current is put into the SQUID and we measure how much current we have to inject to cancel a magnetic field
(From our POV what we’re measuring is the magnetic field next to the person’s head)
What is the approx MEG sampling rate? (At York)
~1000Hz (1000 times per second)
measurements down to 1ms
Physiological MEG artifacts
- Cardiac
- Breathing
- Eye movements
- Muscle movements
Non-physiological MEG artifacts
- Magnetic material in the room
- Cars
- Electronics
- Other magnetic “noise”
- Poorly installed equipment
What is a butterfly plot? (MEG)
Shows magnetic flux/field vs time for all of the channels
What is a sample period? (MEG)
amount of time between samples (1 / sample frequency (Hz))
More generally:
Duration = 1/frequency
What is a nyquist frequency?
Maximum theoretical frequency we can see
in theory: sample rate / 2
- in reality, we limit the signals to a maximum of 200Hz (this is know as our badwidth)
What is frequency resolution?
depends on how much data we have, we can tell our signal granularity
Calculated as: 1/(data length in s)
e.g. 1/10 = 0.1Hz (we can tell differences that are 0.1Hz apart)
the more data the better the resolution
“high” Gamma
70-150Hz:
- cog processing
very high freq oscillations
> 150Hz:
- low level perceptual processing / epilepsy
Distance between the “peaks” of the dipole pattern (in MEG) gives a clue as to…
How deep the dipole is
- Dipole strength drops off as 1/r (cubed) with increasing depth
meaning that deeper dipoles will have a weaker magnetic signal
HOWEVER
- deep focused source will give the same/v similar signal as a more distributed shallow source
What is an Equivalent Current dipole (ECD)
A summation of currents of many neurons with the same positive-negative direction can be mimicked as one strong dipole.
They are comprised of an:
- orientation
- location
- strength (Am)
What is the unit of an ECD (equivalent current dipole)?
Am (Ampere-metres; i.e. current)
Strength of ECD tells us how much current is flowing
From an ECD (equivalent current dipole), we can predict…
what the MEG sensors would show
This is the FORWARD model:
- going from a description of an electrical source (orientation, location, strength) in the head and we predict what it would look like on the MEG sensors
(MEG) evoked (think Event) responses are also known as…
phase-locked responses (or ERF - evoked response fields)
They are the same every time, you can get a few responses and average them starting from the trigger (Like ERPs in EEG)
(MEG) induced responses are also known as…
Non-phase-locked responses
Unlike ERFs (Evoked response fields - equivalent of ERPs from EEG), these can be jittery and misaligned, thus, cannot be averaged
Analysis involves averaging the frequency domain representation:
- create freq(y) time(x) plots for each trial and average them out
Phase-locked (____) responses are a ____ part of the overall MEG signal
(Evoked: think Event), small
Total response = phase locked (evoked) + non-phase-locked (induced)
What does sensor space analysis mean? (MEG)
Looking at the data on the sensors
What is source space analysis? (MEG)
Trying to work out what is happening at a particular location in the brain based on what we observed outside
- Where is something happening
- when is it happening
- what exactly is happening
This is known as the “INVERSE PROBLEM”
- opposite of the forward model (I.e. inferring from a magnetic field what is happening in the neurons)
Dipole modelling is…
the simplest form of inverse problem solution which involves fitting a single dipolar source
- most suitable for simple responses hypothesised to come from a single neuronal source (single location of grouped neurons)
Minimum Norm estimates is a form of … solution
inverse problem
- Like dipole modelling, attempts to explain all of the data on the MEG sensors (at selected times) with a single solution
What is Beamforming
MEG scanning technique
- Takes each source (scans brain) location in turn and estimates activity at that point
- for a given point in space a beamformer is a SPATIAL FILTER which attempts to focus on particular spatial location and attenuate (reduce) responses from other locations
!Simply a weighted sum of sensor activity we see
- once we have the weights for each “location” in the brain, we can produce a “virtual electrode”
What is a “virtual electrode” in MEG?
An estimate of the current (in Am) flow at a point in the brain
What is good about MEG?
- It is a measure directly coupled to brain activity (not via haemodynamics)
- excellent temporal resolution
- Doesn’t suffer from spatial “smearing” of EEG, signals are much more focal (magnetic signal doesn’t smear because of various tissues it travels through)
- Can ask interesting questions about connectivity and networks beyond just “where is this happening”. How do different areas work together(induced and evoked responses)?
Individual MEG sensors can see ___ but not ____ radial sources
tangential, radial
Whether an MEG channel is a magnetometer or gradiometer setup is determined by:
The configuration of the pickup (sensor) coils
Minimum norm is a _____ analysis method
source-space
(EEG) Evoked activity is analysed by…
taking a direct average of the time-series MEG data.