7.1 fMRI 3 Flashcards
(26 cards)
What is the difference between first and second level analysis in fMRI experiments
first level is analysis of an individual and second is analysis of a group of scans (group mean etc)
What image processing is essential to undertake second level analysis?
register all scans in the same space (using a standard space template) -> enables voxel-wise analysis
If there is no fMRI activation in a scan, what is the value of Beta?
Beta/amplitude is zero
For second level analysis, which model is better: mixed effects or fixed effects? why is it better than X model ?
mixed effects as it considers the group variance (between subjects) and the variance of the original MRI data. Fixed only considers latter.
Mixed effect analysis generalises to population as a whole whereas fixed effects analysis is only applicable to the specific subjects studied
What are some examples of higher level analysis settings that you can give FSL to calculate for eg. two different groups? (group a and b)
settings:
show voxels where,
group a activation > b
b> a
mean of group a
mean of group b
What are EVs?
explanatory variables
What would bilateral finger tapping block design experiment involve?
investigating fMRI signal when tapping left and/or right finger maybe vs rest (depends what researcher decides)
What happened in the diabetic pain study?
investigating whether diabetic patients react to the pain they’re subjected to compared to what they have been told (high or low pain stim) -> to test pain response in diabetics
Why is spatial smoothing beneficial?
improves the signal-to-noise ratio
Name some pre-processing images steps
brain extraction, registration, spatial smoothing, motion correction, slice timing, temporal filtering, unwarping,
What are motion correction, slice timing, temporal filtering and unwarping examples of?
Hoe does each of these improve the fMRI image and experiment?
-pre-processing steps
-motion correction = simple rigid body registration is use to ensure consistent anatomical coordinates between images
slice timing = use temporal derivative to ensure consistent timing of slice acquisition
temporal filtering = high pass filtering to remove slow signal drifts (linear ramp thing)
unwarping = corrects for B0 inhomogeneity induced image distortion
What does a brain mask do?
Is a conservative brain mask good?
-brain extraction/skull stripping (non-brain voxels are zeroed)
-conservative is good as you should leave parts which you are unsure about so you can remove them later
How much spatial smoothing must you use in fMRI?
What does spatial smoothing improve? however why is this also bad for the image?
-deciding the amount of spatial smoothing to use is arbitrary in fMRI (although there are FSL calculations to help)
-Improves signal-to-noise ratio at the expense of spatial resolution
What transformation is used in motion correction?
simple rigid body transformations (translation and rotation)
Is slice timing correction used in practice/ when conducting experiments? What is used instead?
not really, FSL uses the temporal derivative as an additional EV in the GLM to account for the small differences in slice timing
What is unwarping?
Why is it used?
-corrects for B0 inhomogeneity induced image distortions
-used as EPI images can be quite distorted, particularly at tissue boundaries
Why is unwarping used in EPI images?
EPI images are prone to distortion
What is slow signal drift? what does it look like?
How is this corrected in a pre-processing step?
-where the waves drift upwards (linear ramp thing)
-temporal filtering: apply a high-pass filter to the data
Unwarping
Why are EPI images prone to distortion?
What is the type of distortion dependent apon in EPI images?
How is this corrected?
-because EPI uses phase-encode ‘blips’ to acquire all kspace data in one TR -> EPI images are susceptible to magnetic field B0 inhomogeneities.
-the direction and polarity of the phase-encode blips can make positive or negative blip distortions
-unwarping = the difference between images with positive and negative blips can be used to estimate a field map and correct for the distortion
Is a rigid body transform sufficient for registration in fMRI experiments?
no! more noisy and complex transforms such as affine or non-linear transforms are needed
For fMRI registration, why dont we use a simple method and directly register the fMRI volume to the standard space template?
What is a better approach to fMRI registration? why?
-because it rarely produces an accurate registration
-using a high resolution STRUCTURAL (eg T1) image from the same subject as an intermediary (because fMRI aren’t usually that high resolution)
What is a orthographic view?
When presenting fMRI data, what is it good and bad for?
-shows fMRI activation in axial, sagittal and coronal planes
-good if you only have a few activation clusters but bad if you have many
If you have a lots of activation clusters, how would you present your results?
What are the advantages and disadvantages of this method?
-light box view in axial coronal and sagittal planes
-gives a good overview of spatial extent of activation but images are small (not good for published papers)
What is the typical way of presenting your results if you only have a few activation clusters?
-orthographic view with significant cluster presentation to accompany
