Session 8 - Advanced fMRI: Functional and effective connectivity

Question 1

functional specialisation

Answer 1

= analyses of regionally specific effects: which areas constitute a neuronal system?

Question 2

functional integration

Answer 2

= analyses of inter-regional effects: what are the interactions between the elements of a given neuronal system?

Question 3

functional connectivity (def)

Answer 3

= temporal correlation between spatially remote neurophysiological events (mechanism free)
= statistical dependencies among spatially remote neurophysiological events

Question 4

effective connectivity (def)

Answer 4

= the influence that the elements of a neuronal system exert over another (mechanistic model)

Question 5

Connectivity analysis

Answer 5

–> asks: how are signals at two different brain locations interrelated?

• use matrix: put all voxels in chain (n_vox, left to right) by time (n_t, up down) –> time series (how are they related?)

Functional connectivity analysis = how are signals at different locations in the brain (inter)related? –> NOT working together, signal each other etc because there is no direct measure, just a correlation

Question 6

Functional connectivity: seedvoxel-based resting state

Answer 6

= extract time series y from a seed region and perform voxel-wise test for correlations

Look at single voxel over time (time sereies y for a seed region):
- fluctuations in magnitude over time
- signal is largely determined by anatomy to very little signal on top (activity) + noise (attention, breathing, blood) in background –> slow fluctuations (usually filtered out)
- frequency transform to get very slow frequencies

Analysis:
- get seed voxel –> central voxel, interested in how this is related to other parts
- get time course in resting state
- check for similar ‘up and down’ pattern in resting state

Issues:
- temporal delay is not accounted for in normal Pearson Correlation –> there are other methods to account for it (e.g. causality approach BUT not unproblematic)

Claim/Interpretation:
- same function? –> could indicate backup system
- work together? –> communication, information exchange by signal transmission
- process same information?
- same background activity? –> no general activity just the same noise

Question 7

What does functional connectivity really mean? Which underlying processes does ‘functional connectivity’ reflect?

Answer 7

• different possible directions and
• it might not reflect connectivity and direction at all: just correlation

1. directed causal influence: A causes B and therefore A and B show a similar pattern
2. bidirectional interaction: A causes B to change and B causes A to change making their patterns more simiar
3. indirect causal influence: A causes C to change which causes B to change
4. common input: C influences both A and B in a similar way

• interrelatedness is not connection directly –> we don’t actually know what it means when regions are
  ‘connected functionally’

Question 8

What happens if voxels and their correlation changes in a task-dependent manner? Psycho-physiological interaction

Answer 8

Psycho-physiological interaction (PPI):
= identify brain regions that exhibit a task-dependent chnage in correlation with a seed region
- attention versus no attention to visual stimulus –> does it amplify the relationship between the regions (Which area receives increased input from V1 under attention vs no attention?)

What to do?
- multiply task with brain acitivity: 1 is on, -1 is off –> build a regressor that has properties of being positive if there is more correlation and negative with less correlation
- multiply with 1 when task and -1 when no task –> get enhanced/inversed picture

But:
- difficult interpretation, at network-level we can’t really say what happens
- problem cannot be solved at level of original time series

Question 9

Dynamic causal modelling

Answer 9

• solve problem in a model-based way: which model makes better predictions?

Neuromodel and fMRI level model: haemodynamic model
- decompose neural model: direct extrinsic input (all that comes from the outside) –> goes to V1, baseline connectivity
- corresponding locations in V1 and V2 (point to point) and within-layer, task-dependent changes in connectivity

Issues:
- theories are not at level of fMRI responses
- modelling, not measuring: within-layer connectivity changes (with attention?), between-layer connecitivty (feedforward, feedback)
–> formal model comparisons needed: which works better?
–> be careful about interpretation: model versus measure