MEG, fMRI

Question 1

intro to fmri

Answer 1

• It works by placing a participant into a strong, static magnetic field, generated by a large superconducting electromagnet, cooled by liquid helium (needed to obtain superconductance)
• The field strength is usually 3 tesla (T) for experimental research (but there are also stronger magnets – University of Melbourne has a 7T scanner), or 1.5 T for
  clinical purposes (the earth magnetic field is 65microtesla)
• Participants are placed into the scanner, and their head is covered by a coil (RF coil)
• There are also gradient coils, which are used to modify the (otherwise ideally homogenous) magnetic field for short periods of time

Question 2

safety of fmri

Answer 2

• Being exposed to a strong magnetic field is harmless to participants (some people do it hundreds of times)
• Moving in the field can make some people nauseous
• However, there is the danger of attracting magnetic objects (metal) which can be fatal – and since the field itself is invisible, it is easy to forget that it exists

Question 3

participant view

Answer 3

• Participants can view experiments, which are controlled from outside the scanner room, via mirrors mounted on the head coil, or via goggles
• Responses can be given via scanner-compatible keys, joysticks, or a touchpad
• Participants’ head position is fixed to avoid any movement, which would distort the signal
• The head coil is used to send radio frequency (RF) pulses and also functions as a receiver for the incoming signal

Question 4

Magnetic Resonance Imaging (MRI) – The Basics

Answer 4

In order to understand how functionalMRI works, we first need to get a basic understanding of how a structural image is acquired using MRI

• For this, we have to cover some of the basic physics of nuclear magnetic resonance (NMR), which is another word for the same method (but it does not refer to the use of any radioactive materials!)
• NMR refers to the anatomic nucleus, which contains protons and neutrons
• More than 70% of the human brain consists of water, which contains Hydrogen atoms (H+ protons)
• These can be thought as small bar magnets, “precessing” like a spinning top about an axis

Question 5

Hydrogen in fmri

Answer 5

• The Hydrogen atoms are often simply referred to as protons, or “spins”
• Their “precessing” is often referred to as “spinning”
• The protons’ spin directions are initially random, but in a strong, externally applied magnetic field, like in the MRI scanner, they align parallel or anti-parallel to the magnetic field (often referred to as B0)
• Most protons align parallel to the B0 field
• However, they are not perfectly aligned – and they are also not static, but they still keep “precessing” in a random fashion

Question 6

B0 field

Answer 6

• The B0 field is oriented into the direction of the Z-axis in the scanners coordinate system
• The precession frequency of the protons is also referred to as Larmor frequency
• Importantly, the precession frequency of protons depends on the strength of the magnetic field. This means, we know precisely the frequency with which they “spin” because we know how strong the magnetic field is.
• For simplicity, we can now imagine all protons being aligned with the B0 field, but they would all be in different positions in their precession (i.e. the phase in
  which they spin is different)
• However, as long as they are aligned in the direction of the B0 field we cannot measure a signal with our head coil, which surrounds the head
• The second problem is that the signal from each proton itself is tiny, and they are not precessing in phase (i.e. they are not in the same position at the same time)

Question 7

radio frequency

Answer 7

In order to get a signal, we apply a radio frequency (RF) pulse perpendicular to magnetic field B0 using the head coil
- If the frequency of the RF pulse matches the precession (Larmor) frequency of the protons, it will affect these protons

Question 8

effect of rf

Answer 8

• The first effect of the RF pulse is that all protons will start precessing (or spinning) in phase, meaning that their magnetisation will all point to same location in space at the same time
• This happens because the protons absorb energy from the RF pulse (which also heats up the tissue a bit)
• The second effect of the RF pulse is that the magnetisation vector (i.e. the net magnetisation that the protons have together) is tilted away from the Z-axis
• This means, the magnetisation is tilted from the longitudinal direction (of the field B0) into the transversal plane (the X-Y-plane)

Question 9

receiving signal from fmri after rf

Answer 9

• The magnetisation vector now “rotates” around in the transversal plane where our head coil is placed
• The head coil will now receive this as a signal
• However, this signal is still rather meaningless for us, because it comes from the entire brain
• The trick is now to now switch off the RF pulse after which the transversal magnetisation decays very quickly because the protons emit the excess energy
• They also lose phase coherence very quickly, which makes the signal disappear
• Finally, the original longitudinal magnetisation will recover

Question 10

T1 and T2

Answer 10

• The first consequence of switching off the RF pulse is that protons align again with the magnetic field, also referred to as longitudinal relaxation, spin-lattice relaxation (because it is due to an interaction of “spins”, the protons, and “lattice” the local environment), or T1 recovery
• T1 refers to a time constant of a function, indicating how long the recovery of the longitudinal magnetisation takes
• Importantly for us, this time constant is different for different tissue types
  The second consequence of switching off the RF pulse is that the transversal magnetisation decays (which is an independent process), also referred to as transversal relaxation, spin-spin relaxation (because it is due to an interaction of “spins”, the protons, with other protons), or T2 decay
• T2 is the time constant indicating how long the transversal decay takes
• T2 decay is much faster than T1 recovery, and again different for different tissues

Question 11

Getting image from fmri via decay

Answer 11

• When the signal is measured during this phase of relaxation, different signals will be emitted from protons in different tissues
• Depending on when signals are measured, researchers can use T1 and T2 (and the density of protons) to get differently weighted images of the brain and clearly see the type of tissue (because the signal will be differently strong)
  For this, different types of sequences (e.g., T1-weighted or T2-weighted) are used that are optimised to capture differences in signal, due to T1 recovery or T2 decay (or proton density)
• Structural brain images depend on when signal is recorded during this process

Question 12

Reconstructing brain images (slice 1)

Answer 12

• In order to get separate measurements from different locations in the brain, we first need to reconstruct where exactly the signal comes from
• For this, we use gradients created by the gradient coils
• Protons will absorb energy from RF pulses only when the frequency of the RF pulse matches the proton’s precession (also called “resonance”) frequency
• Thus, by causing the magnetic field to vary linearly, we can cause the resonance frequency to vary throughout the brain
• An RF pulse of a specific frequency will now only excite one slice of the brain –precisely the slice where the resonance frequency of the protons matches the frequency of the RF pulse

Question 13

slice selecting gradient

Answer 13

• The first gradient we use is called slice selecting gradient
• It varies the gradient field along the Z-axis such that different slices are exposed to different field strengths
• A RF pulse can now be chosen to match precisely the precession frequency of protons in one “slice” of the brain
• This gradient is applied during the RF pulse

Question 14

phase encoding gradient

Answer 14

• Now that we know the “slice” the signal comes from, the second gradient we use is called the phase encoding gradient
• It changes the precession (or, “spin resonance”) frequency of the excited protons depending on their location in the gradient, causing de-phasing
• When removed, the resonance frequencies are the same again, but the differences in phase persist – their phase is now informative about their position
• This gradient is applied after the RF pulse

Question 15

frequency encoding gradient

Answer 15

• The third gradient is called to frequency encoding gradient, and it changes the magnetic field within the selected slice
• This happens during read-out of the signal
• Because all protons at a certain position in the gradient now have same resonance/precession frequency, the frequency at read-out is informative about their position

Question 16

Gradient reversals

Answer 16

Gradient reversals are used to “un- do” the effects of the gradients

• The final signal consists of a series of “echos” elicited by the reversals (hence, sequences that use this trick are called gradient-eco, GE, sequences)
• Reversals of the RF pulse can also be used to create an echo (used in spin-echo sequences)
• This entire process, repeated for each slice, takes some time, meaning that it takes 1-3 seconds to measure the entire brain once

Question 17

Other important parameters are in fmri

Answer 17

• In classical sequences, one RF pulse is used for each slice, meaning that the time it takes to measure the entire brain depends on the number of slices (called TR, time to repeat)
• The other important parameter is the echo time (TE), which is the interval between excitation and data acquisition
• Other important parameters are:
  oThe slice thickness and gaps between them
  oThe size of a measurement point, i.e. the “voxel”, the tree-dimensional pixel
  oThe field of view (FOV), i.e. how many voxels we measure per slice
  oThe number of slices we want
  oThe orientation of the slices and read-out direction

Question 18

the beginning Hemodynamic imaging

Answer 18

• fMRI does not measure neural activity directly but is a hemodynamic neuroimaging method
• Seiji Ogawa discovered in the early 1990s that large blood vessels cause “brighter areas” (i.e. better signal) in MRI scans
• Ogawa’s group (and another research team around Robert Turner) investigated this phenomenon in detail
• Ogawa changed the blood-oxygen level experimentally and found that this indeed impacts the signal

Question 19

The science behind hemodynamic

Answer 19

• Local neural activity requires energy, which is generated in the brain in the form of adenosine triphosphate (ATP)
• To produce ATP, glucose is metabolised, for which oxygen is required
• Oxygen is constantly transported through the brain arteries and a network of arterioles (small arteries)
• In the capillaries, oxygen molecules are removed from hemoglobin (Hb), turning oxyhemoglobin into deoxyhemoglobin
• Deoxygenated hemoglobin is then transported away by the venules and the larger veins

Question 20

What happens when there is local neural activity talk about blood oxy

Answer 20

• First, there is a slightly delayed increase of glucose and oxygen consumption
• This triggers an increase in cerebral blood flow (and blood volume) to supply more oxygen
• The consequence is a local increase in blood oxygenation
• The increase in blood oxygenation is much larger that the initial dip, meaning that shortly after the neural activity, there is an oversupply of oxygen in the blood
• The increase in blood oxygenation causes our signal to get better – this is the Blood Oxygen-Level Dependent (BOLD) signal we measure
• Oxygenated blood (oxyhemoglobin) is diamagnetic, enhancing the signal

Question 21

Hemodynamic Response Function

HRF

Answer 21

• The change in signal is described by the Hemodynamic Response Function (HRF), which is similar (but not identical!) in different brain regions
• The peak of the HRF is reached 4-8 seconds after the neural activity occurred
• It takes the signal up to 16 seconds to go back to baseline levels
• If several neural events take place, their HRFs will add up linearly

Question 22

What is reflected by the BOLD signal

Answer 22

• The relationship between the BOLD signal and “neural activity” is complicated, but that should not be a reason to refrain from using it
• The BOLD signal correlates best with local field potentials (LFPs), which reflect the neurons’ “input” at the synapses, but recent studies have shown that it is also correlates with action potentials (i.e. “neural firing”; the “output” of neurons)
• In sum, the BOLD signal might be a mixture of activity within local cortical excitation-inhibition networks (EIN), small and highly interconnected functional microunits, which show massive recurrent feedback

Question 23

From measuring the BOLD signal to getting “results”

Answer 23

• Due to the presence of deoxygenated hemoglobin, which is paramagnetic (i.e. it has a “magnetic momentum”), protons (hydrogen atoms) experience an additional speeding up of the decay of transversal magnetisation
• This means, the total decay is faster than predicted by the T2 constant
• The total decay (or “de-phasing”) is referred to as T2* decay. BOLD fMRI mostly uses T2*-weighted pulse sequences (e.g., Gradient-echo echo-planar imaging, GE EPI), which give a better signal when blood is more oxygenated
• In a typical fMRI experiment, the HRF can only be measured sparsely
• The HRF is must therefore be estimated, and timing relative to the event of interest matters

Question 24

Statistical Parametric Mapping

Answer 24

-For analysis of BOLD we use Statistical Parametric mapping
Using a General Linear Modelling (GLM) approach, we search for regions in which the signal increase fits to the predictions of our model
- The program is given the mapping between conditions and recorded brain images and then estimates the model fit
- Irrelevant variables, which might impact the estimate of the model, can also be factored in, but are not analysed
- This is done first for each voxel in each participant’s data
- Group statistical analyses are then performed on the sample

Question 25

What happen after a result statistical map of BOLD is created

Answer 25

• The resulting statistical maps (indicating how well the data fit the model) is then overlaid on a structural image of the brain
• Significant effects in a brain region for task A compared to a control task B (or the baseline) is interpreted as involvement (or “activation”) of this region
• We cannot compare activation between regions because the HRF is different
  =Activation “blobs” are statistical effects in experiment, often colour-coded for “activation” (red) and “de-activation” (blue)

Question 26

Repeating of BOLD

Answer 26

• We need to repeat the measurement of brain activity many times while participants perform the experimental task because the signal is very noisy, and the final brain maps are always based on averaging across many trials
• But now we understand the entire chain from experimental task to brain “results”

Question 27

Are the significant “blobs” really the (full set of) brain regions that compute
the cognitive task

Answer 27

• Because we apply strict statistical tests to avoid interpreting “false positives” (i.e.statistically significant results, which are in reality just random), we risk overlooking real results (i.e. we might only see the “tip of the iceberg”)
• “Better safe than sorry” approach
• We have >50,000 voxels in the brain, and we run a t test for every voxel.
• Even if we use a p < .01 value for each test, the risk of rejecting the null hypothesis but being wrong (a “false positive”) is still 1% - for each test!
• In this example we would expect 500 (!) false positives – that is, significant voxels even when in reality there was no real difference between conditions in any voxel

Question 28

Method to avoid too many false negative for bold

Answer 28

to avoid risk of too many false positive
The strictest correction is Bonferroni-correction: divide the significance level (e.g., 0.01) by the number of tests
= voxels), and use this new significance level for each test: 0.01 / 50,000 èp < .0000002(corresponds to the overall risk of 1% to have one false positive)

Question 29

If we find activation “blobs” in the brain, does that mean we found the
“module” for the function of interest

Answer 29

• Cognitive functions are often realized by inter-connected brain networks, not by single regions: “[…] a unified mind has no components to speak of” (Logothetis, 2008)
• There are methods to investigate other aspect of brain function with fMRI, for example the connectivity between regions

Question 30

Can we see the entire cognitive process unfolding in our brain images with high temporal resolution

Answer 30

• Because it takes ~1-2 seconds to measure the entire brain once, we cannot see any changes that take place within this time period
• This is often referred to as the “poor temporal resolution” of fMRI
• If we want to measure neural activity changes for fast processes, we need a different method (such as EEG or MEG)

Question 31

Can we see all neural activity there is in our brain image

Answer 31

• The smallest measurement unit is a “voxel”, which is a 3D pixel, and the standard voxel size is ~ 3 x 3 x 3 mm
• This means, we can learn nothing about what happens within a voxel
• However, since a single voxel still contains >100,000 neurons, there will be a lot going on that we will miss

Question 32

Understanding face processing

Answer 32

• One of the classic fMRI studies by Kanwisher and colleagues (1997) attempted to investigate how faces are represented in the brain
• They presented their participants with images and measured the BOLD signal. They then contrasted face processing to the processing of objects: faces > objects

Question 33

Kanwisher et a on face processing

Answer 33

• Kanwisher et al. found a region located in the fusiform gyrus responding more strongly to faces than to objects
• They could show this result reliably in most of their participants, and they could replicate it with different participants

Question 34

Control of kanwhiser et al

Answer 34

• To rule out that this result was simply due to using objects as a control category, they replicated the study with the contrast faces > scrambled faces
  Kanwisher, 1997 Contrast: faces > scrambled faces
• This analysis confirmed the initial findings of stronger activation for faces in the fusiform gyrus
• They also used another category of objects for the contrast faces > houses (also worked)
  Finally, they reasoned that it might be useful to contrast faces with different other body parts, so they used the contrast faces > hands
• Again, this analysis confirmed the initial findings of stronger activation for faces in the fusiform gyrus

Question 35

rep ofKanwisher et a

Answer 35

This finding has been replicated many times, and it is one of the strongest in fMRI research
- They even named the region “fusiform face area (FFA)

Question 36

result of KAnwisher on module

Answer 36

• Kanwisher assumed that they found the brain’s “module” for face processing
• Since then, some other “modules” have been discovered:
  oFusiform Face Area (FFA)
  oParahippocampal Place Area (PPA): houses and places
  oExtrastriate Body Part Area (EBA)
  oSome regions specialised for letters / tools / animals
• But not all existing objects seem to have their own “module” in the brain…

Question 37

FFA and PPA rep

Answer 37

• Kanwisher’s results have been well replicated, in particular the FFA and PPA
• These regions reliably activate even when people imagine faces and houses

Question 38

main argue point agaisnt kabwisher

Answer 38

• But others were not convinced that Kanwisher was right…
• One argument is that it would be simply impossible if the brain had “modules” for all possible objects – the existing “modules” already take up a lot of space
• We would run out of space and just couldn’t represent everything we know
• And how could we represent new objects that we don’t know yet?

Question 39

Is the FFA really doing face processing (the greeble)

Answer 39

• Gauthier and colleagues also measured face processing. In addition, they asked their participants to distinguish between “Greebles” – strange and totally novel (and faceless!) objects participants had never seen before
• During the experiment, participants learned the family structures and became experts for Greebles

Question 40

Result of greeble study

Answer 40

• First, when participants did not know much about Greebles, the FFA responded strongly to faces, but not to Greebles
• This is the result that would be predicted by Kanwisher’s studies
• However, after becoming a “Greebles expert” the FFA also responded strongly to Greebles, not just to faces
• This means, activation in FFA might reflect expertise Gauthier, 1999
• Everyone just happens to be an expert for faces
• Other studies have since confirmed the activation of the FFA for other areas of expertise (e.g., butterflies, chess configurations)

Question 41

The eye interpretation of FFA

Answer 41

• Another radically different interpretation of what the FFA does is that it simply represents what is typically found in the centre of our vision
• In this view, the FFA is specialised for objects that require a high resolution, meaning that they are processed foveally
• Faces just happen to be in the centre, requiring high resolution vision, all the time

Question 42

Malach and colleague exp (visual eye)

Answer 42

• Malach and colleagues argue that the organisational principle in the ventral visual cortex (i.e. the part of the brain that processes objects) might follow cortical topography, i.e. an eccentricity mapping
• The visual system might not be organised by specific object categories, but by where in our visual field objects are usually encountered
• Places/houses are represented in the PPA not because this is the module for places, but this is where the visual system processes objects that are usually encountered in the periphery of our visual field

Question 43

Evidence for all three?

Answer 43

• Research found evidence for all three coding schemes – and suggested that all of them might be true to some extent
• The fMRI signal might therefore reflect a mixture of different coding schemes
• This is a great example of how difficult it is to interpret fMRI results – even if we first think that there is clearly just one possible explanation

Question 44

How are objects encoded in the brain

Answer 44

Haxby and colleagues found that the brain uses a distributed code to represent many different object categories

• They argued that in order to represent all possible objects, objects must be represented in a distributed fashion
• This means that all objects are represented in the entire “object region” in the brain, not only in specialised modules
• Based on the idea that the brain uses a distributed code to represent all possible objects, researchers have build classifiers to learn these codes from fMRI results
• It is now possible to predict which object a person is seeing (from all object pictures on Google) from patterns of fMRI activity with very high accuracy

Question 45

Reverse inference

Answer 45

We “draw conclusions about cognitive processes from the presence of activation”
– this is called reverse inference, because we already need to know what the activation means in order to draw these conclusions

Question 46

Problem with inference example

Answer 46

(1) In this study, when task A, then brain region Z is active
(2) In other studies, when cognitive process X, brain region Z is active
(3) Thus, in this study: activity in Z èengagement of cognitive process X
- The problem is that assumption (2) is not exclusive: brain region Z may be active for many other tasks, not just A, so it might also be related to other cognitive processes, not just X

Question 47

Infering the frontal cortex

Answer 47

• For example, the regions in prefrontal cortex are notoriously difficult to “understand”
• These regions appear to be activated by many different cognitive tasks
• Duncan (2010) proposed that these prefrontal regions are part of a Multiple Demand Network, which computes many high-level cognitive processes
• Rather than serving one particular function each, they might be recruited more strongly the more demand there is (i.e. the more difficult the task is)

Question 48

Anti duncan op

Answer 48

• Others are not as pessimistic about being able to learn from observing activation in these areas and suggest that there might be specialisation, not so much about the task, but about how abstract the cognitive process is
• More anterior regions (towards the front of the brain) represent more abstract information
• More posterior regions (towards the back) represent more specific content

Question 49

The ending of frontal infer

Answer 49

• It remains highly debated who is correct; however Duncan now accepts that
  there is some specialisation in the prefrontal cortex
• This means, he thinks that there are relatively more neurons in sub-regions recruited by different tasks (or cognitive processes), but there is no absolute specialisation
• If the task is difficult enough, then neurons from more and more prefrontal regions are recruited

Question 50

Measurement of cognitive processes exp

Answer 50

• “If the experimental setup fails to manipulate the cognitive process of interest, it cannot provide useful information about that process”
• Relatedly, if task A involves multiple cognitive processes,we do not really
  know how much it tells us about cognitive process X
• In consequence, if we observe specific brain region Z being activated, we still do not learn much about the cognitive process X

Question 51

Problems when using fMRI to learn about cognition : what does it depend on

Answer 51

Poldrack expressed these problems in probabilistic terms. The probability that we really learn from our fMRI results that cognitive process X is involved depends on:
oThe quality of the task to measure this particular cognitive process
oThe specificity of the brain region to reflect only this cognitive process

Question 52

how to interpret “null findings”

Answer 52

• If a contrast (e.g., task A > task B) was not significant in a particular brain region, does that mean that this region is not involved in the cognitive process?
• Our statistical tests are designed to make it difficult for the H1, not for the H0, meaning that we can’t interpret null results
• Our fMRI method might also simply not be sensitive enough to detect the differences between conditions, even if they exist
• For example, in a primary visual region, if we contrast vertical lines vs. horizontal lines, we might not see any significant results
• The reason is that in each voxels, there are many neurons coding for all possible line orientations (colour-coded in the figure)
• This is simply due to the spatial resolution of fMRI. With a typical voxel size of 3 mm3, we have clusters of neurons (called “cortical columns”), each “firing”
  for one orientation, all in one voxel
• If we had smaller voxels (without losing signal quality), we might indeed be able to detect differences

Question 53

General issues with interpreting fMRI data (conclusion on causal effect)

Answer 53

• The consequence is that we can never conclude that brain regions are notinvolved in a cognitive process
• Unfortunately, research papers still often imply this Neuroimaging often seems to add value to our results (i.e. a better explanation), but that’s not always the case, so always ask what you really learn from it
• If you want to do more than just localising a cognitive function, you need to come up with very clever experimental paradigms
• All methods – fMRI, EEG and TMS – have their strengths and weaknesses, and they are useful for answering different questions that require different spatial and temporal resolution