Dear Anthony,
Your colleague is wrong I'm afraid.
The MNI152lin is the result of doing an average of 152 subjects using
linear registration. This was generated many years ago.
The MNI152 is the result of doing an average of the same 152 subjects
using a much newer non-linear registration method. It is fairly new.
Note that it is still blurry in areas which have substantial variation
in
anatomy, such as in the parietal areas.
There is also a "Colin" brain which is the average of many scans of
the same individual (Colin) but we do not distribute this image in
$FSLDIR/data/standard.
Your colleague is correct in saying that it is better to use a
population
average image, but both MNI152 and MNI152lin are population averages.
Our recommendation is to use the MNI152 with our non-linear registration
as this tends to give the best results, since the image is sharper and
the
non-linear registration can take advantage of that.
All the best,
Mark
On 28 Oct 2008, at 23:23, Anthony Ang wrote:
> Hi Jesper,
>
> Thank you very much for your very good advice. It has cleared my
> doubts.
>
> Yes, the reason why I am thinking of using MNI152_1mm is because my
> subject image is of 1x1x1 mm voxel resolution so I thought that I
> should also use a 1x1x1 mm voxel MNI152 image (something like
> comparing an apple to an apple, rather than a pear) - I have now
> understood this thinking is apparently wrong. I have checked my
> warped image (MNI152 warped to my subject image) and it is 1x1x1 mm
> voxel.
>
> Please further pardon my ignorance. What's the difference between
> MNI152 and MNI152lin? In what circumstances to use MNI152 or
> MNI152lin?
> MNI152 is very clear while MNI152lin is blurry. A fellow colleague
> said that I should use MNI152lin because it is an average of 152
> brains (that's why it is blurry) and not MNI152, which is multiple
> scans from a single brain (Colin). He said since I have 60 different
> brain images (3 scans of 20 subjects each), I should use the
> MNI152lin as it represents an average of the population. My question
> is does it really matter if I use MNI152 or MNI152lin?
>
> Thanks.
>
> Best regards
> Anthony
>
>
>
> On Tue, Oct 28, 2008 at 7:40 PM, Jesper Andersson <[log in to unmask]
> > wrote:
> Hi again Antony,
>
> >
> > If I wish to use MNI152lin_T1_1mm image as the reference file,
> then I will
> > need to create a T1_2_MNIlin152_1mm.cnf file?
> > In this file,
> > --ref=MNI152lin_T1_1mm
> > What will be the -refmask?
>
> Just a piece of advice here. Why do you want to use the 1mm template
> instead? If you are estimating the warps with a resolution of ~8-10mm
> (i.e. --warpres=10,10,10) the 1mm template is overkill and it will
> take
> much longer to run without improving your results.
>
> If the reason is that you want to do your final resampling to 1mm
> (i.e.
> have your finished images at 1mm resolution) you can do that anyway.
> Just
> do
>
> fnirt --ref=MNI152_T1_2mm --in=my_ima --cout=my_warps --config=MNI...
> applywarp --ref=MNI152_T1_1mm --in=my_ima --warp=my_warps
> --out=my_warped_ima_at_1mm
>
> Good luck Jesper
>
|