If you have coregistered T1 weighted images, then maybe estimate the
normalisation parameters from these. Otherwise, you could maybe look for
clues about why normalising via segmentation failed, by looking at the tissue
class images that get generated. I suspect that a segmentation model that
assumes that brains consist only of GM, WM and CSF, and that these tissue
types have fairly uniform intensity, is not likely to work so well on these
FLAIR images.
Best regards,
-John
> I am trying to normalize T2 FLAIR images of patients with brain lesions.
> First, I set the origin (AC) of the images on SPM5¡¯s display command.
> Then, I ran SPM5¡¯s unified segmentation process to get normalization
> parameter file. I used the default setting except for the warping
> regularization value, which was set to 30 because the normalized images
> were severely distorted when I set the value 1.
>
> I normalized the images with the parameter file obtained from the
> segmentation process, with default setting only changing bounding boxes to
> show entire brains. After normalization, some of the images were very
> tilted and out of correct origins. So I corrected the tilts and set the
> origins again on SPM¡¯s display command, and reoriented and resized the
> images.
>
> But still, normalized images don¡¯t seem to be normalized into the same
> stereotaxic space, because they show different locations on the same slice
> number. I¡¯ve also tried standard SPM5 normalization process with lesion
> masking, but it also led to the same result.
> Attached files are same slices of the normalized images of different
> subjects.
>
> Was there something that I did wrong?
> What do you suggest I do? Please help me.
|