UM_STACK02 functions (for the structural images)
1. The first step is to convert the files from the GE header format to
an Analyze single file. The output file name is something like
"t1overlay.img"
2. The image is now filtered to remove the dielectric effect (this
means that we get larger flip angles and therefore different signal
intensity at the center of the image). We use Kalina Kristoff's spm_homocor.m. The image name is now "et1overlay.img"
3. The images are now converted to NIFTI format. The image name is now "ht1overlay.nii"
4. The skull is removed to facilitate the normalization process (some
times the images are acquisred with fat saturation, other times they
are not. To work around the discrepancies between images and the
standard templates we simply remove the scalp from the images and from
teh template. There is a scalped template here ). The scalping is done usinf BET from the FSL library. The image name is now "eht1overlay.nii"
---------
UM-STREAM7 functions (for the BOLD time series):
1. The first thing the script does is to build a directory structure
for storing the data. The data are stored in the following
directory structure for a given subject whose ID is "subjectID":
./subjectID/func
./subjectID/func/run_01
./subjectID/func/run_02
.... etc. |
This is where the functional BOLD weighted data goes.
They are organized by run ( a run is defined as a time series of
images collected at a constant TR without stopping the scanner).
Each run directory contains the initial reconstructed images,
physio corrected images, slice timing corrected images, realigned
images. |
| ./subjectID/phys |
This is where the physiological data are stored in
the cases where it is collected (typically in functional connectivity
analyses) |
| ./subjectID/anatomy |
This is where the structural images are stored.
Typically this means T1 weighted overlay images and T1 weighted
high resolution SPGR images. |
| ./subjectID/raw |
The raw k-space files ("Pfiles") are stored here.
If anything goes wrong, this is where we go back to in order to
rebuild the data. |
2. The raw Pfiles are filtered to remove "white pixel artefact" if
present. (White pixel artefacts are large spikes in the k-space
data that lead to a stripe pattern superimposed on the reconstructed
images).
3. The data are reconstructed into images. The reconstruction
incorporates field map correction to reduce susceptibility
artefacts. (more info here soon). The output file is called
"run_01.nii" for the first run, "run_02.nii" for the second ...etc.
4. If physiological data are collected, they are used at this stage to
remove physiological artefacts from the time course using the RETROICOR
algorithm (Hu et al: Magn Reson Med 34, 201-212 (1995).
Pfeuffer et al Magn Reson Med 47, 344-353 (2002)). The
output file name is now "prun_01.nii".
5. The next step is to correct the images for the difference in the
acquisition time of each slice. Typically the slices are
collected sequentially from bottom to top, so the bottom slice was
collected almost one full TR before the top slice. In this step,
we shift them in time to align them with the middle slice.
We use "slicetimer" from the FSL library . The output file name is now "aprun_01.nii"
6. The images are next realigned. The stream looks for the 10th image in the first run for reference. We use "mcflirt" from the FSL library with the following parameters:
mcflirt -in tmp -out rtmp -refvol 0 -cost normcorr -verbose 1 -stats -plots -mats
The output file name is now "raprun_01.nii"
----
Note that the output of all these steps is in NIFTI format.
In a nutshell, the whole time series are stored in a single file
sequentially, and a header (348 or 352 bytes long) is pre-pended to the
file. The images names end in .nii. Here is a matlab program to convert from NIFTI to 4D Analyze format.
Click here
for more information on How to normalize
Back to the Lab
|
|
