Resting State fMRI

Question 1

What is the basis of fMRI?

Answer 1

The basis of fMRI is during a task you have activated neurons when then increases blood flow, it pushes out oxygenated blood which causes deoxygenated blood to decay and generates the MRI signal we measure

Question 2

What type of blood is diamagnetic?

Answer 2

Oxygenated blood

Question 3

What type of blood is paramagnetic?

Answer 3

Deoxygenated blood

Question 4

How does the brain consume energy?

Answer 4

Weight of the brain is 2% of body mass, yet consumes 20% of energy

In terms of BOLD activity, research has found that BOLD only actually increases 1-2% from baseline

Question 5

What happens during resting state which is also considered a limitation?

Answer 5

Spontaneous low-frequency neuronal activity that also produces BOLD signal

Brain is not ever ‘resting’ fully, uncertainty in what the brain is doing in its ‘uncontrolled’ state casts doubt over what is being explicitly recorded as various processing may simultaneously occur during ‘rest’

Question 6

Why should we study the brain at rest?

Answer 6

Assessing task-based localisation vs resting state tells us a lot about localisation and connectivity- how brain regions are connected, function & communicate- aligns with the core aims of neuroscience as a principle

Potential to be a clinical and cognitive biomarker
- Can give insight into many disorders, e.g. schizophrenia found to related to global disconnectivty in the brain

Can be done in any population e.g. babies or elderly or diseases - does not require an additonal demands as its baseline activity

Quick setup with relatively little expertise

Question 7

What was the first paper to discuss resting state fMRI?

Answer 7

Biswal et al (1995)

Discovery of functional connectivity

Task condition-finger tapping
Rest condition- doing nothing

Extracted the time-series from the left motor cortex and correlated the signal with every other voxels’ time-series (whole-brain analysis)

Found a strong correlation with the time-series of the opposite hemisphere’s motor cortex - suggesting that these two functionally similar regions generated similar patterns of activity even at rest.

Question 8

What did Shulman et al. (1997) find?

Answer 8

Anti-correlational PET study- found two networks of brain regions and showed inverse pattern in blood flow depending on whether a task is performed or not

Did verbal and non-verbal task - found that certain regions e.g. anterior cingulate and sulcus were consistently active regardless of task

Summarised that the blood pattern of activation maps in this region were task independent

Question 9

What are three types of connectivity in fMRI?

Answer 9

Functional – how are different regions in the brain functionally connected

Structural- how are different regions in the brain structurally connected

Effective- what is the influence of one brain region to another

Question 10

What are resting state networks?

Answer 10

Networks consistently found in a lot of research

Collectively called the Beckmann Eight Networks

Question 11

List the Beckmann Eight Networks.

Answer 11

Default mode
Right fronto-parietal attention
Left fronto-parietal atttention
Executive control
Medial visual cortical areas
Lateral visual cortical aras
Auditive system
Sensorimotor cortex

Question 12

What are the two main features of resting state data?

Answer 12

1. Resting state data is very replicable- same networks have been independently verified by De Luca et al. (2006), Damoiseaux et al. (2006) and others
2. Resting state is very close to task-based activation patterns

Question 13

What is one of the most important questions for resting state fMRI work?

Answer 13

Is it real?
Is the signal neuronal in nature, does the fluctuation really reflect synchronised neuronal fluctuation?

When you say two brain regions are functionally connected, you assume the signal between them goes up and down together (synchronised fluctuations) but there are also addtional factors that contribute to the MRI signal, e.g. noise, motion, scanner instability, image constuction errors

Question 14

Is resting state signal neuronal in nature?

Answer 14

Yes - Cordes et al (2001)

Frequencies contributing to functional connectivity in the cerebral cortex in “resting-state” data

Findings:
Low frequency band corresponds to neuronal activity
High frequency signals corresponds to arteries etc - high frequencies overlap with respiration and cardiac frequencies as well

Conclude: Functional connectivity in the auditory, visual, and sensorimotor cortices is characterized predominantly by frequencies slower than those in the cardiac and respiratory cycles. In functionally connected regions, these low frequencies are characterized by a high degree of temporal coherence.

They summarise its highly unlikely the resting state signal is only driven by heart beat or respiration

Question 15

What does evidence from Laufs et al. (2003) suggest about resting state being neuronal in nature?

Answer 15

Conducted an EEG resting state study and examined correlation between EEG power and resting state fluctuations

Found very good regional correlation between fluctuations in the resting state fMRI signal and fluctuations in the power of EEG

Question 16

How long is a resting state scan?

Answer 16

6 minutes - all they do is look at a cross and try not to think of anything

Question 17

What are the typical parameters for resting state acquisition?

Answer 17

6-10 mins is usually enough
T2* weighted pulse sequence
Tr=35ms
TE= 2600ms
Continual trans-axial slices of 4mm thickness
In-plane pixel dimension 1.8mm x 1.8mm

Question 18

What software is used for RS analyses?

Answer 18

FSL
CONN Toolbox (SPM)

Question 19

What is a typical resting state analysis pipeline?

Answer 19

Aquire fMRI data
-then goes through pre-processing
-then goes through de-noising stept
-then goes through post processing
-then choose network measures or predefined measures to generate the maps you want

Question 20

What is pre processing in RS data?

Answer 20

SImilar to task-based
Get a reference image
Firstly you re-align the data and co-register it to a structural or T1 scan
Then apply slice timing and detect your outliers
Then you normalise your data
Then smooth it

Question 21

What is an important step in resting state data?

Answer 21

Denoising

Noise signals might be: head-movement, mean white-matter/CSF signals, cardiac/respiratory/ETCO2 measures, others.

Question 22

What is denoising?

Answer 22

Noise’ can be filtered/regressed out of the data prior to analysis

OR

Noise signals can be entered into the analysis models to explain additional variance

Question 23

How do we denoise?

Answer 23

ICA-based methods
Mean-signal based methods
Linear regression and temporal band filtering

Question 24

What is the denoising setup?

Answer 24

Conn toolbox for resting state
Do linear regression, removie confounding regressors e.g. CSF, white matter, motion etc

Once you’ve done this the value should be centred towards 0 which means you’ve done a good job getting those confound out

In the toolbox you can also do the skew checks

Question 25

What are some methods you can use in resting state analysis?

Answer 25

Voxel-based methods- cluster of results
Node-based methods- nodal results

Question 26

Give some examples of voxel-based methods.

Answer 26

Seed-based correlation analysis
Independent component analysis

Question 27

Give an example of a node-based method.

Answer 27

ROI to ROI
Network modelling analysis

Question 28

What is an independent component analysis?

Answer 28

Exploratory “model free” or data driven approach
Multivariate voxel based approach
Finds interesting strucutres in the data
Its a spatial approach oftne called global brain approach as we find patterns in the whole brain

ICA is a group of tools that allows you to decompose signals into high or low frequencies (in this case)

Question 29

What is spatial ICA?

Answer 29

X = A * S

fMRI data = time course x spatial maps

fMRI data measured (X) is expressed as a set of unknown spatial patterns (S) and the associated time courses (A)

Question 30

What is the ICA functional connectivity pipeline?

Answer 30

Single subject - identify interested networks - extract connectivity measures- group condition/comparison

Group- identify interest networks - dual regression - group condition/comparison

Question 31

What is groupwise ICA?

Answer 31

Expressed as before X = A * S

X is now all the subjects (concatonated subjects)

Instead of a single ICA you get a group ICA spatial map

Important as analysis tool to study specific conditions within or between groups of subjects

Question 32

How does groupwise ICA work?

Answer 32

Set up in toolbox as before
Select to do a GICA
Select 30 spatial components (what most people use)
Choose the parameters e.g. which type of dual regression to use

Question 33

What is the issue with 30 spatial components?

Answer 33

Its hard to tease out the networks in 30 spatial components

Question 34

How can we tease out networks in the toolbox?

Answer 34

In the toolbox you have a spatial template which gives you the probability of which spatial components from the GICA matches the Beckmann Eight Networks

Question 35

What is dual regression?

Answer 35

1st stage = subject specific time courses
2nd stage = subject specific spatial maps

1. Regresses the group-spatial-maps into each subject’s 4D dataset to give a set of timecourses
2. Regresses those timecourses into the same 4D dataset to get a subject-specific set of spatial maps

It is then common to compare the spatial maps across groups of subjects to look for group differences

Question 36

What is seed based connectivity data analysis?

Answer 36

Calculate correlation coefficient between two areas or regions
Localised approach
User specifies the seed or ROI of interest

Examines two timecourses from two different areas and calculates a correlation coefficient

Find the connectivity of a seed to the rest of the brain. The seed can be a collection of points based on prior fMRI studies or based on an atlas, in which case it is a seed with a larger region-of-interest (ROI).

Question 37

What is the seed based connectivity pipeline?

Answer 37

Define ROI or seed region - extract time series- extract connectivity measures - group condition/comparison

Question 38

What is ROI to ROI analysis?

Answer 38

ROI-to-ROI connectivity metrics characterize the connectivity between all pairs of ROIs among a pre-defined set of regions.

Question 39

What is the ROI to ROI setup?

Answer 39

Initial setup same as seed-based but for second level setup you can select areas central to the research you’re doing e.g. for neuropathy patients you could select the somatosensory areas, thalamus, insula etc, just areas linked to the research
You can also restrict the analysis

Question 40

What is the ROI to ROI functional connectivity pipeline?

Answer 40

Define ROI or seed region - extract time series - extract connectivity measure - group condition/comparison

different to seed-based:
When you extract the connectivity matrix, instead of it being an ROI to every voxel, its actaully a ROI to a ROI so its a nodal approach
You get all of this and map out the connectivty matrix and then you do a GLM or compare the group or condition in the second level and what you get is a node, a significant region to target