Fiber Tracking & DTI

Question 1

What does structural connectivity refer to?

Answer 1

It is physical white matter fibre connections between remote brain areas

Question 2

What is functional connectivity?

Answer 2

It is statistical dependencies in neural activation time series of remote brain areas

Question 3

What is the difference between structural and functional connectivity in a nutshell?

Answer 3

How are physically connected vs. How these connections are put into use

Question 4

What technique is used to measure structure connectivity?

Answer 4

Diffusion Tensor Imaging (DTI), a specific MRI sequence

Question 5

Who coined a term connectome?

Answer 5

Sporns and Hagman did it independently in 2005
A connectome (connection + -ome) is a network of structurally or functionally connected brain areas

Question 6

How is connectivity matrix of the brain interpreted?

Answer 6

Columns and rows correspond to brain areas, the cells tell us whether the elements are connected

Question 7

What directions do colours in tractography maps code?

Answer 7

Dorsal-ventral - blue
Posterior-anterior - green
Left-right - red

Question 8

What brain tissue is in focus of DTI?

Answer 8

White matter (inside fibre tracks, axonal connections) — connectome is about the connections between these matters

Question 9

What is diffusion?

Answer 9

It is a property of molecules to distribute evenly in the space (based on Brownian motion)

Question 10

Why can diffusion of water molecules in brain be informative about fibre tracks?

Answer 10

Fat and water don’t go well together — diffusion is not happening through the fat, it is happening along the fat
Fat in the brain = myelin
Diffusion of water molecules in the brain depends on local myelin content

Question 11

What is a tensor?

Answer 11

A diffusion tensor is a mathematical representation of the diffusion properties of water molecules, an item used to describe diffusion

Question 12

How are isotropic and anisotropic diffusion and, hence, tensors different?

Answer 12

Isotropic
Diffusion: no external restriction => happening in all directions similarly, e.g. ventricles
Tensor: ball-like, with similar eigenvectors
Anisotropic
Diffusion:only in parallel to the axonal fibres, never in perpendicular, e.g. corpus calossum
Tensor: cigar-shaped, one eigenvector is more stretched into a certain direction

Question 13

What is fractional anisotropy? How is it measured?

Answer 13

Fractional anisotropy (FA) is a scalar value derived from diffusion tensor imaging (DTI) that measures the degree of anisotropy or directionality of water diffusion.
FA values range from 0 to 1, where 0 is isotropic diffusion (equal diffusion in all directions) and 1 is highly anisotropic diffusion (diffusion predominantly along a single direction).

Question 14

How is the starting voxel called?

Answer 14

A seed-voxel

Question 15

What is a streamline?

Answer 15

The term streamline is used to designate the contiguous set of 3D points produced by tractography algorithms

We cannot claim that we are actually tracking fibres, that is why the term streamline is used

Question 16

What anatomical contraints are there for the streamlines?

Answer 16

• Start: in grey matter — seed voxel at GM/WM
• End: in grey matter — crop at GM/WM
• Should never go through CSF — exclude implausible fibre tracts

Question 17

Describe the connectome workflow

Answer 17

1. High-resolution structural MRI (to delineate GM and WM) + DWIs (for WM)
2. Different brain areas are defined on the basis of cortical thickness maps
3. Superimposing brain areas on DTI and looking at the results of fibre tracking — whether there are streamlines that touch our ROIs
4. Subtract the streamlines that are not touched
5. Connectivity matrices: when we don’t have directional information, the matrix is symmetrical: e.g. there is more connectivity within the hemispheres than between the hemispheres, mostly connected via cortical matter
6. An alternative to a matrix is a circular plot: two hemispheres

Question 18

How much individual variations are there in brain connections?

Answer 18

A lot of individual variation:
* only 7% of possible connections are present in all participants;
* 20% of possible connections are not present in any participants

Question 19

What are two approaches to representing connections in a network?

Answer 19

Binary vs. Weighted connections = Connection present/absent vs. Scalar

Question 20

How to weigh connectivity between two areas?

Answer 20

1. Number of streamlines
2. Length of fibres
3. (Average) fractional anisotropy: 0-1
4. Generalised fractional anisotropy: also takes into account additional diffusion properties, such as radial diffusivity
5. Streamline volume density: the density or concentration of streamlines (fiber tracts) within a specific volume of brain tissue, dissimilates between strong connections and connections between big areas
6. Streamline surface density: the density of streamlines on the surface of brain structures or ROIs

Question 21

How is a DTI acquired?

Answer 21

• Diffusion MTI is structural MRI (anatomy information, not functional connections => the data is no affected by any cognitive processes)
• Works on all modern MRI scanners (even though there are scanners specifically optimised for the connectome)
• At least 6 (= 3 directions multiplied by 2) diffusion directions needed (more are beneficial, e.g. 98)
• Time to acquire one full scan ~15 minutes
• Procedure: RF with 90 flip angle -> diffusion gradient -> RF with 180 flip angle -> diffusion gradient; signal is acquired in 90ms between FID and echo

Question 22

What is a graph?

Answer 22

A graph is mathematical representation of a network G=(V,E), it is a collection if vertices (nodes) and their relationships (edges)

Question 23

What are the characteristics of brain networks?

Answer 23

• Most networks have low network degree — exponential degree/strength distribution => high resilience of the whole system, low chance of destroying the network by randomly lesioning a random network (heavy-tailed distribution)
• Brain paths (a number of edges you need to pass through to get to a node — short paths in the brain): the brain is rather sparse, there are not so good connections between clusters
• A high degree of local clustering: 3 nodes, if 2 are connected, is the 3rd also connected — short paths and very efficient connectivity within a cluster
• Modularity: a lot of connections inside a module, a few connections outside a module

Question 24

What are two defining features a small world network? What is common and what is different from regular and random networks?

Answer 24

A small-world network is a type of network characterized by two key properties: high clustering coefficient and short average path length.

Regular networks: strong connections between immediate neighbors, but long paths
Random networks: low clustering, but short paths

Question 25

What measure is used to assess the “small-worldness”? Who introduced it?

Answer 25

Small-World Propensity is a measure that quantifies the extent to which a network displays small-world structure (0-1, 0.6> — small-world network)
It was introduced by Bassett and Bullmore

Question 26

What is a hub in a network?

Answer 26

Hub is a highly connected node or region that plays a crucial role in facilitating communication and integration between different parts of the network

Question 27

What is the principle of a rich club?

Answer 27

Brain regions with the most connections preferably connect with each other

Question 28

What connections are there with respect to the rich club?

Answer 28

• Rich club connections
• Local (peripheral)
• Feeder (between local and rich club)

Question 29

Why are rich club connections called the backbone of the brain?

Answer 29

• Short paths usually involve local connections, but any long path goes through the rich club
• when rich clubbers are taken out, the communicability (how well two brain areas communicate), drastically drops, with feeders not that much, with local insignificantly

Question 30

Rich club connections may be a key to what phenomena?

Answer 30

• The rich club may be an answer to the comorbidity of some diseases
• There is also evidence that in schizophrenia there is less connectivity in rich club, while feeders and local connections are not affected

Question 31

Do structure and function correlate?

Answer 31

Yes, structural and functional connectivity of one person do actually correlate.
Horn et al., 2014:
- DTI and resting-state fMRI, 40000 row-wise correlations
- highest correspondence between voxel-wise FC and SC in the default mode network => DMN is the functional brain network, which uses the most direct structural connections