Transcript Image Preprocessing for Idiots

Overview
fMRI time-series
Motion
correction
kernel
Design matrix
Smoothing
General Linear Model
Spatial
normalisation
Statistical Parametric Map
Parameter Estimates
Standard
template
PHJ
a) Direct Normalization
i) Realign -> Slice Time* -> Normalization ->
Smoothing
b) Indirect Normalization
i) Realign -> Slice Time* -> Coregistration > Segmentation -> Normalization >Smoothing
* optional
Lars Kasper
Realignment
fMRI time-series
 Aligns all volumes of all runs
spatially
 Rigid-body transformation:
three translations, three
rotations
 Objective function: mean
REALIGN
squared error of
corresponding voxel
intensities
 Voxel correspondence via
Interpolation
Motion corrected
Mean
functional
Signal, Noise and Preprocessing
3
Realignment Output:
Parameters
Signal, Noise and Preprocessing
Lars Kasper
4
The preprocessing sequence revisted
Realignment
– Motion correction: Adjust for movement between
slices
Coregistration
– Overlay structural and functional images: Link
functional scans to anatomical scan
Normalisation
– Warp images to fit to a standard template brain
Smoothing
– To increase signal-to-noise ratio
Extras (optional)
– Slice timing correction; unwarping
Co-registration
Term co-registration applies to any
method for aligning images
– By this token, motion correction is also coregistration
However, term is usually used to refer to
alignment of images from different
modalities. E.g.:
– Low resolution T2* fMRI scan (EPI image) to
high resolution, T1, structural image from the
same individual
Co-registration: Principles behind
this step of processing
When several images of the same
participants have been acquired, it is
useful to have them all in register
Image registration involves estimating a
set of parameters describing a spatial
transformation that ‘best ‘ matches the
images together
fMRI to structural
Matching the functional image
to the structural image
– Overlaying activation on
individual anatomy
– Better spatial image for
normalisation
Two significant differences
between co-registering to
structural scans and motion
correction
– When co-registering to
structural, the images do not
have the same signal intensity
in the same areas; they
cannot be subtracted
– They may not be the same
shape
Problem: Images are different
Differences in signal intensity between the
images
Normalise to appropriate template (EPI to EPI; T1 to T1), then segment
Segmentation
Use the gray/white estimates from the
normalisation step as starting estimates of the
probability of each voxel being grey or white
matter
Estimate the mean and variance of the
gray/white matter signal intensities
Reassign probabilities for voxels on basis of
– Probability map from template
– Signal intensity and distributions of intensity for
gray/white matter
Iterate until there is a good fit
Register segmented images
Grey/white/CSF probability images for EPI
(T2*) and T1
Combined least squares match
(simultaneously) of gray/white/CSF
images of EPI (T2*) + T1 segmented
images
The preprocessing sequence revisted
Realignment
– Motion correction: Adjust for movement between
slices
Coregistration
– Overlay structural and functional images: Link
functional scans to anatomical scan
Normalisation
– Warp images to fit to a standard template brain
Smoothing
– To increase signal-to-noise ratio
Extras (optional)
– Slice timing correction; unwarping
Normalisation
Goal: Register images from different participants
into roughly the same co-ordinate system (where the
co-ordinate system is defined by a template image)
This enables:
– Signal averaging across
participants:
Derive group statistics -> generalise
findings to population
Identify commonalities and differences
between groups (e.g., patient vs.
healthy)
– Report results in standard coordinate system (e.g. Talairach and
Tournoux stereotactic space)
Matthew Brett
Standard spaces
The Talairach Atlas
The MNI/ICBM AVG152 Template
The MNI template follows the convention of T&T, but doesn’t match the particular brain
Recommended reading: http://imaging.mrc-cbu.cam.ac.uk/imaging/MniTalairach
SPM: Spatial Normalisation
SPM adopts a two-stage procedure to determine
a transformation that minimises the sum of
squared differences between images:
Step 1: Linear transformation (12-parameter affine)
Step 2: Non-linear transformation (warping)
High-dimensionality problem
The affine and warping transformations are
constrained within an empirical Bayesian
framework (i.e., using prior knowledge of the
variability of head shape and size): “maximum a
posteriori” (MAP) estimates of the registration
parameters
Step 1: Affine Transformation
Determines the
optimum 12-parameter
affine transformation to
match the size and
position of the images
12 parameters = 3
translations and 3
rotations (rigid-body) +
3 shears and 3 zooms
Rotation
Shear
Translation
Zoom
Step 2: Non-linear Registration
Assumes prior approximate
registration with 12-parameter
affine step
Modelled by linear
combinations of smooth
discrete cosine basis functions
(3D)
Choice of basis functions
depend on distribution of
warps likely to be required
For speed and simplicity, uses
a “small” number of
parameters (~1000)
Matthew Brett
2-D visualisation
(horizontal and vertical
deformations):
Ashburner; HBF Chap 3
Brain
visualisation:
Source
Deformation
field
Template
Warped
image
Smoothing
 Why blurring the data?
 Improves spatial overlap by blurring over




anatomical differences
Suppresses thermal noise (averaging)
Increases sensitivity to effects of similar
scale to kernel (matched filter theorem)
Makes data more normally distributed
(central limit theorem)
Reduces the effective number of multiple
comparisons
Kernel
SMOOTH
 How is it implemented?
 Convolution with a 3D Gaussian
kernel, of specified full-width at halfmaximum (FWHM) in mm
Signal, Noise and
Preprocessing
22
MNI Space
GLM
Lars Kasper: A Toolbox
Smoothing
Example of
Gaussian smoothing in
one-dimension
The Gaussian kernel is
separable we can smooth
2D data with two 1D
convolutions.
Generalisation to 3D is
simple and efficient
A 2D
Gaussian
Kernel
23