Lecture 11: Spatial accuracy and resolution of MEG and MCQ Flashcards
The spatial accuracy of MEG source localisation can be pretty good - (4)
Zimmerman et al., (2019) compared MRI and MEG
localisation of M1 (primary motor cortex) in a glioblastoma patient
Both results shown here on the structural MRI scan
Found similar areas/locations
Despite being difficult due to reorganisation after tumour
Why is spatial accuracy of MEG of source localisation is better than EG
electrical signals are dispersed by the tissue they pass and how much varies by the type of tissue, and volume currents are the main signal picked up
An amazing recent study showed spatial accuracy of MEG for source localisation is good for beamformers - (4)
Within 1 mm under ideal conditions in brain map to actual correct source
Used monkey brain
Using optogenetics in the monkeys brain (injecting a chemical that makes a portion of cells respond to light -acting like retina) so that activity occured at an exactly known time and place
Can detect subcortical activity (at least in small monkey brains) –> ideal as small brain and less complex task –.> harder for big brain and complex tasks
So, what is the spatial resolution of MEG? Is it millimetres?
- might not be mm due to human head moving and distributed network
The spatial resolution in MEG depends on many factors such as - (7)
- Source location – varies around the brain!
- Source orientation
- Accuracy and type of head model
- Estimating for one condition vs. between conditions
- Signal-to-noise ratio
- Number of sensors
- Specific methods used
What does this diagram show? - (4)
- stimulated data on red and red dot is exact area
- Finding the exact source location in great and good
- In third case, , the activity is deep in the insula but its saying its across the temporal lobe its because its a deeper source its struggle in locating the soruce
- Finding right source location in some areas and wrong source location in other areas using inverse modelling in stimulated data –> due to leakage
Leakage in MEG is
effect of other locations on the activity at a specific location
What does this diagram show in terms of leakage? - (6)
- Using stimulated data using forward and inverse model say we have activity at this green dot and activity around it
- Because we have so much sensors like 250 and estimating 15,0000 sources
- The sources we get are not independent (related across sensors) and weighted versions of the different sensors
- Because of that we get at green dot is made up of activity happening at all these regions
- Activity at green dot affects all these regions
- If we are effecting something cms away,we can’t say our spatial accuracy/resolution is mm - realistically its cms (better than EEG) and lower than fMRI.
Fourier analysis breaks down the MEG signal into what?
A. Sine waves of different events
B. Lego bricks of different colours
C. Sine waves of different amplitudes
D. Sine waves of different freuqencies
D
Univarate analyses look at the activity changr at each sensor/vertex
What do multivariate analyses look at?
A. The activity change across all sensors/vertices
B. The differences in activity between sensors/vertices
C. The pattern of activity over sensors/vertices
D. The freuency of the activity over sensors/vertices.
C.
Time-freuency
When responses are time and phase-locked, frequency based analyses are best - T or F
When responses are not time-locked, event-related analyses are best - T or F
When responses are time but not phase locked, time-frequency analyses are best - T or F - (3)
F
F - when time locked do event-related analysis
T
We use an individual’s brain mesh to decide where their dipoles (sources) should go
Where do we typically put them?
A. deep in brain mesh in line with the cortes
B. on the cortical surface pointing towards it
C. On the cortical surface of the cortex in line with it
D. Deep in the brain mesh pointing towards the cortex
B - sources on the cortical surface as won’t get deep down as that is where pyramidal cells is and pointing towards (sinc epyramidial cells perpendicular towards cortical surface)
The forward model solves the forward problem by transforming?
A. activity in sensor space into source space
B. into a giant rock
C. activity in source space into sensor space
C
True or false
MNE uses spatial filters to look at each location independently
Distributed SL methods like dipole modelling produce an actiivty map
Beamformers struggle to find correlated sources
F - false because beamformers use spatial filters to look at each location independently, MNE looks at all locations and make smooth map
F - False because distributed methods produce activity map but dipole modelling is not a distributed method
T - Trying to find one and exclude everywhere else but if its correlated missing true effect
Activity at which source would have the LEAST effect on the data at this sensor
A. Red
B. Orange
C. Purple
C - purple since its not going towards the sensor
