Dear Ben,

hope you are fine with me CC'ing the mailbase. It may be useful for others too.

I had a look at your data, and it registers acceptably well. Both the functional->structural and the structural->standard. The structural->standard is not perfect in the topmost slices. It is made difficult by signal from the meningies (I think it is) that has almost identical intensity as the grey matter. But it is still acceptable.

I notice that the structural scan you sent me was named CNMC0194-msr1_brain.nii.gz, but that is still contained the scalp. If you want to use fnirt in feat you need to supply two versions of your structural scan.

1. A betted version whos name ends in _brain
2. A non-betted version with an identical name to the one above but which does not end in _brain

I suspect your non-betted scan with the ending _brain might have been used by flirt as if it was betted, giving very poor results. These will then be fed into fnirt which will then also fail.

So try that and let me know how it goes.

Good Luck Jesper

On 6/3/10, Jesper Andersson <[log in to unmask]> wrote:
Dear Ben,

I would need to properly see your data to be able to help you with this. Could you please upload the structural scan you are using as well as the example_func file to

and then send the six-digit number to my mail address (not the mailbase).


On 2 Jun 2010, at 20:19, Ben Yerys wrote:

Hi FSLers--

Thanks to Jesper and others for input on my registration woes.  Removing the neck from the MPRAGE was one of several big problems in registering the High Res image to the Standard.  However, I'm still having significant problems with registration.  I do have a fair amount of ghosting in my images, because Fat Sat was turned off in order to have a sequence with TR=2000.  At least I was told that Fat Sat had to be turned  off for the sequence to be this fast.

I've attached one picture which highlights my problem.  It appears as though the EPI image is highly distorted in the right hemisphere (left side of display) and my thought is that this is driving all of my registration problems.  I ran BET ("apply to 4D") with the default fractional intensity threshold of 0.5 to my 4D data and this removed almost all of the ghosting except for the two most superior slices and the distortion was reduced but still a problem.  My other pre-processing steps include:
Slicetiming *I've taken this out in case I made a mistake in my slicetiming file
Smooth FWHM
Registration - Main structural image - Linear: Normal search with 6 DOF
Registration - Standard Space - Linear+Nonlinear with MNI152_T1_2mm_brain Warp Res=10mm

The participant is a typically developing 12-yr-old who had a max translation movement of 0.5mm.  I ran this data through SPM5 and from what I can see by eye using check reg, the registration looks acceptable and my activation map is in line with predictions for my set-shifting task.  What am I missing?

Thanks for the help!
<highres_to_standard_neck removed3.jpg>

Benjamin Yerys, PhD
Assistant Professor
Center for Autism Spectrum Disorders
Children's Research Institute - Neuroscience
Children's National Medical Center
111 Michigan Ave, NW
Washington DC, 20010
202-476-5358 (office)
301-765-5425 (lab)