HC7 fMRI Flashcards
To measure the resting state of fMRI..
Participants are instructed to rest and think of nothing
particular. Only 1 or 2 scans are made every 8 minutes. It is very feasible.
Methods to get a better view of the amount of data and to better interpretate the data
- Principal component analysis (PCA)
- Independent component analysis (ICA)
- Graph theory
Principal component analysis (PCA)
PCA identifies key components that capture most of the variance in the data. Each component represents a group of brain regions with highly correlated activity, allowing for data reduction while preserving important patterns.
Independent component analysis (ICA)
The data is a mix of signals. You have to identify statistically independent components (1 for each independent signal). Every component reflects an independent signal. Regions with the same time course, belong to the same source, so this is a way to separate the data. (see picture on the right). → blind source separation method.
Graph theory
Gives you parameters that summarize network properties. It calculates the connectivity between two regions. It is used for functional connectivity and resting connectivity. If the functional connectivity with other regions is high, the node has a high degree.
Resting state networks
- Default mode network
- Somatomotor network
- Visual network
- Language network
- Dorsal attention network
- Ventral attention network
- Frontoparietal control netwok
How were those networks found?
All these networks are active even though they are at rest. These networks are found by using the independent component analysis.
What is the main limitation of fMRI in the future?
fMRI can only evolve with better spatial resolution because the HRF (Hemodynamic Response Function) is a biological process that does not change, causing low temporal resolution.
What is voxel-wise ( or univariate) analysis?
You look at every voxel and threat it individually. The voxel-wise or univariate in which nearby voxels are treated individually or expected to show similar signals (spatial smoothing).
What is Multi-voxel (or multivariate) analysis?
We are interested in the differences between voxels and a pattern of activity between voxels. For example, when we have a neural activity of 9 voxels in 2 conditions with the same overall activation, we investigate whether the pattern of activation across voxels is different between conditions. The level of activity can be the same for the two conditions, but the activation pattern of the voxels can differ. See picture below.
What does multi-voxel pattern analysis (MVPA) investigate?
It looks at whether the pattern of voxel activation differs between conditions, even if overall activation levels are the same.
What are the two types of multi-voxel analysis?
Correlational MVPA
Decoding MVPA
Correlational MVPA
Tests whether within-condition correlation between datasets is higher than between-condition correlation. If so, the region differentiates between conditions. This is the easiest approach.
Decoding MVPA
Decoding MVPA uses voxel activity patterns from one dataset to train a classifier, which learns to distinguish between conditions. The classifier identifies a decision boundary in a multi-dimensional space, without needing to explicitly define it. If a new pattern falls on one side of the boundary, it belongs to category one; if it falls on the other side, it belongs to category two.
What is the purpose of decoding MVPA?
It trains a pattern classifier to classify neural activity into different categories based on voxel activation patterns.
What is cross-validation in decoding MVPA?
Using a new dataset to find the same test classifier. So it is replicable.
How does decoding MVPA relate to free will?
Researchers could predict which button a participant would press before they were aware of their intention, questioning the concept of free will.
How can decoding MVPA be used in ASD (autism) research?
It helps classify subgroups of ASD patients using MRI-based biomarkers in brain regions.
What are the advantages of MVPA?
- Detects differences between conditions that univariate analysis cannot.
- Measures the magnitude of differences in voxel activation patterns.
What is fMRI adaptation?
A neural response decreases when a stimulus is repeated, indicating neuronal selectivity in a voxel.
What assumption is made in fMRI adaptation?
Areas with a diminished response to repeated stimuli are sensitive to stimulus features.
Why does the fMRI signal sometimes decrease for different stimuli?
Some neurons activated by the first stimulus also respond to the second, causing partial overlap and slight adaptation, there is a little adaptation and thus the fMRI signal decreases a little bit.
What are the three models of fMRI adaptation?
Fatigue model
Sharpening model
Facilitation model
Fatige model
There is less overall activation when you sum everything up. All neurons which were involved the first time, will be involved the second time, but to a lesser degree. → lower BOLD response.
Scharpening model
When you sum everything up, only the sharp signals will be active so a fewer neurons are active. All neurons that were highly active the first time, will still be highly active the second time, but the neurons which were a little active won’t be active anymore.
Facilitation model
The processing goes faster, so there is less processing time. The level of activation stays the same. → lower BOLD signal.
Which fMRI adaptation model has the most evidence?
The fatigue model has the most support.
How does attention affect fMRI adaptation?
More attention leads to greater adaptation, but this effect was not observed in Americans in one study.
