On Mon, Oct 26, 2009 at 7:18 AM, Jesper Andersson <[log in to unmask]> wrote:

Dear Veronica,

it is a little hard to known exactly what is going wrong. In general fnirt can sometimes be quite sensitive to data that has been collected with a much shorter echo-time than what the data constituting the template was. Which is jargon for images with much higher intensity in the scrappy bits outside the brain (meninges and stuff). This is something we are aware of and are working to fix.

My suspicion is that that is what you are seeing, caused by the high intensities in the meninges just outside the visual cortex. If you want to you can download the data to our server and I can take a look at it (http://www.fmrib.ox.ac.uk/cgi-bin/upload.cgi).

Jesper

On 24 Oct 2009, at 09:20, Veronica W. wrote:

Hello,

I have been trying to use FLIRT followed by FNIRT to map a patient MRI
to the standard MNI 152 brain. This has produced distorted results, and I
was hoping someone could help me troubleshoot.

In specific, the distortion is that cortical gyri and sulci seem to be
unusually shaped and do not match with the MNI_152 template. There
are extra gyri in some locations and missing gyri in others. Also, there
is an outward distortion at the top of the skull.

Please see below for links to images, my code, and configurations.

I thought the trouble might be the regularization settings, so I have tried
a "small lambda" and "large lambda" configuration as well as using
the standard configuration file. These configurations are listed below (and
results for each are linked). The standard configuration file seems to
produce the best results.

Thanks in advance for any advice you might have!

-Veronica W.
Neuroscience Statistics Research Lab, MIT


##### LINKS TO IMAGES ###########

1. My original MRI
http://web.mit.edu/vsw/www/FSL/Original_MRI.jpg

2. Template MNI 152 2mm brain
http://web.mit.edu/vsw/www/FSL/MNI_2mm_template.jpg

3. Results using T1_2_MNI152_2mm.cnf
http://web.mit.edu/vsw/www/FSL/Conversion_with_standard_2mm_cnf.jpg

4. Results with large lambda config.
http://web.mit.edu/vsw/www/FSL/Conversion_with_high_lambda.jpg

5. Results with small lambda config.
http://web.mit.edu/vsw/www/FSL/Conversion_with_low_lambda.jpg

##### CONVERSION SCRIPT #########

This is exactly as suggested on the FNIRT Example Uses page for
"Registering T1-structural to MNI152"

#1. Perform BET on MRI image
bet $MRI $MRI_BETTED -f 0.5 -R

#2. Use FLIRT to map betted MRI to MNI152_2mm_brain template
flirt -in $MRI_BETTED -ref MNI152_T1_2mm_brain.nii.gz -out
$MRI_FLIRTED_BETTED -omat $MRI_FLIRTED_BETTED_OMAT -searchrx -180 180
-searchry -180 180 -searchrz -180 180 -dof 12 -interp trilinear -bins 256
-cost corratio

#3. Use FNIRT on raw MRI, using as an input the affine transformation matrix
from Step 2
fnirt --in=$MRI --aff=$MRI_FLIRTED_BETTED_OMAT --cout=$FNIRT_OUT
--config=$FNIRT_CNF_FILE

#4 Apply warp
applywarp --ref=MNI152_T1_2mm.nii.gz --in=$MRI --warp=$FNIRT_OUT
--out=$MRI_FNIRTED

####### THREE CONFIGURATIONS I TRIED ############

1. Standard 2mm cnf configuration file
 completely unchanged from T1_2_MNI152_2mm.cnf

2. Large lambda
--ref=MNI152_T1_2mm.nii.gz
--refmask=MNI152_T1_2mm_brain_mask_dil.nii.gz
--imprefm=1
--impinm=1
--imprefval=0
--impinval=0
--subsamp=4,2,1,1
--miter=5,5,5,5
--infwhm=6,4,2,2
--reffwhm=4,2,0,0
--lambda=300,200,100,100
--estint=1,1,1,1
--applyrefmask=1,1,1,1
--applyinmask=1
--warpres=10,10,10
--ssqlambda=1
--regmod=bending_energy
--intmod=global_non_linear_with_bias
--intorder=5
--biasres=50,50,50
--biaslambda=10000
--refderiv=0

3. Small lambda
Same as above, with --lambda=300,75,30,10


Any advice about how to produce less distortion in the cortical gyri
and sulci would be appreciated!  Thank you.