Transcript Lesson 13

M.Tech. (CS), Semester III, Course B50
Functional Brain Signal
Processing: EEG & fMRI
Lesson 13
Kaushik Majumdar
Indian Statistical Institute
Bangalore Center
[email protected]
Poldrack et al., 2011
Different MRI Image Types
Poldrack et al., 2011
Flow Chart of fMRI Processing
Steps
Spatial normalization in case
of group analysis of fMRI
Spatial Smoothing: Filtering out
High-Frequency Components


Removal of high-frequency components
enhances SNR at the larger spatial scale.
Most fMRI analyses are performed across
multiple neighboring voxels.
Noisy acquisition in smaller voxels can be
smoothed out by spatial smoothing
(performed, for example, by convolution with
a suitable window function).
Spatial Smoothing (cont)


During group analysis of fMRI data spatial
smoothing helps even out small individual
differences, which interfere with the general
(group) trend to be studied. All of these are not
taken care of in usual spatial normalization.
Some analysis methods (like, Gaussian random
field) require smoothing.
Amount of Spatial Smoothing
Spatial smoothing is often achieved by
convolution with a Gaussian kernel function
with standard deviation σ. In that case the
amount of spatial smoothing is “Full width at
half maximum” (FWHM) = σ√(2ln2) = 2.55σ.
Also FWHM = √(FWHMintrinsic2 + FWHMqpplied2).
Poldrack et al., 2011
Effect of Smoothing with Different
Applied FWHM Values
Spatial Normalization or
Intersubject Registration


There is considerable variation in minute detail,
shape and size of the brain across individuals.
In order to locate functional activities to specific
regions of the brain, irrespective of individual
differences, intersubject 3D fMR image
registration need to be performed. This is called
spatial normalization.
See for detail Chapter 4 of Poldrack et al.,
2011.
Poldrack et al., 2011
Talairach Coordinate
http://ja.m.wikipedia.org/wiki/%E3%83%95%E3%82%A1%E3%82%A4%E3%83%AB:Gr
ay726_central_sulcus.svg
Anatomical Landmarks
Automated Registration



MNI305 template – created by anatomical
registration of 305 brains in Talairach atlas and
then taking the average across all 305 brains.
MNI305 is the most widely used template in use
today. Activities of a brain under study are
directly mapped on this template.
This template is based on white Caucasian
brains and therefore not ideal in shape and size
for many other brains, such as south-east Asian
brains.
Poldrack et al., 2011
Spatial Normalization Steps
Poldrack et al., 2011
Parametric Transformations
References

R. A. Poldrack, J. A. Mumford and T. E.
Nichols, Handbook of Functional MRI Data
Analysis, Cambridge University Press,
Cambridge, New York, 2011.
THANK YOU
This lecture is available at http://www.isibang.ac.in/~kaushik