[VoxBo] normalization using SPM procedures

[Charan Ranganath, Ph.D.]

> I normalized my functional image(s) [dimension 64*64*30] using EPI.nii 
> template [dimension 91*109*91] in SPM5. But then the normalized image 
> has a dimension of 79*95*69, when my common sense expects the dimension 
> of the normalized data to be the same as the template. And in fact some 
> edges of the brain are cut off. Would this be something normal?
IMHO, MRI analysis software rarely follows common sense. The issue 
probably has to do with the bounding box setting. That, along with the 
resliced voxel size, will determine the new dimensions of the normalized 
image. If the bounding box is too small you may get edges cut off. You 
can change that setting easily in SPM5. I can't tell you what to try, 
but you can experiment and see what works best.

> I also noticed that if I want to normalize a .tes file with 150 time 
> points, then I first convert .tes into 150 different .img files and 
> transfer them to SPM, then normalize, then possibly convert them back 
> into a .tes file. Would this be a normal procedure? It seems like a LOT 
> of grunt work (for all runs of every subject), and thought there's gotta 
> be a way around it.
> 
> I've looked into SPM manuals for these issues but couldn't find a good 
> explanation. Thanks in advance. JP
We get our data in analyze format from the recon software, run it 
through SPM5 up to the GLM phase, then convert to .tes files, run the 
glms, and then make analyze files for the randomfx analysis that we run 
in SPM. It is a bit clunky, but it's not a lot of work, and it runs 
pretty well.

Good luck, CR

> 
> 
> 
> Quoting Charan Ranganath <cranganath at ucdavis.edu>:
> 
>> Hi, I think I can suggest two potential solutions:
>> (1) Take the t1.nii image and skull strip it using the brain 
>> extraction tool (BET). The program is described at: 
>> http://www.fmrib.ox.ac.uk/analysis/research/bet/
>>
>> (2) Normalize your EPIs to the EPI template included in SPM.
>>
>> We have found that we get better across-participant registration with 
>> approach #2. However, if your EPIs have a lot of dropout and 
>> distortion or if they only cover part of the brain, you will have to 
>> use #1.
>>
>> Good luck. CR
>>
>> Joonkoo Park wrote:
>>> Hello,
>>>
>>> As recommended by VoxBo wiki, I'm trying to normalize images using SPM5.
>>> I first convert the (anatomical) .cub files into .hdr and .img (using 
>>> vb2img) and use SPM procedures to calculate *_sn.mat and eventually 
>>> warp the original .img image to a normalized one. I used T1.nii file 
>>> in spm/template folder as a template for calculating parameters for 
>>> anatomical images.
>>>
>>> One problem I see here (well, it would be the same case if I were to 
>>> use VoxBo norm) is that my anatomical images (i guess functional 
>>> images too) do not include skull, but the template includes skull and 
>>> other parts of the head. I am worried that my anatomical image will 
>>> assume that even the skull part of the template is part of the brain, 
>>> and this seems to the case when I look at the normalized images.
>>>
>>> The skull part of my images are eliminated during the preprocessing 
>>> procedure at the scanner site. Is there a way to solve this problem?
>>>
>>> Thanks!
>>>
>>
>> -- 
>> Charan Ranganath, Ph.D.
>> Associate Professor
>> Center for Neuroscience and Dept. of Psychology
>> University of California at Davis
>> 1544 Newton Ct.
>> Davis, CA 95616
>>
>> phone: 530-757-8750
>> fax: 530-757-8640
>>
>> http://DynamicMemoryLab.org
>>
>>
>>
>>
> 
>