If the DICOM import of SPM works for your particular flavour of DICOM,
then you should not need to do any re-orienting of images imported using
this facility (unless the subject was lying in some very uncomfortable
orientation in the scanner).  The converter in SPM reads the DICOM
header fields to figure out the appropriate image orientations.  Other
converters may not do this (in which case, there is a high risk of left
and right sides of the brain becoming confused).

If you only intend to use the data in SPM5/8 (or any other NIfTI
compliant software), then no further rotating should be done (your flips
actually encode a rotation - as the determinant of the matrix is still
positive).

Best regards,
-John

On Wed, 2010-03-03 at 18:31 +0800, 黄沛钰 wrote:
> IMHO, I think there're several stages for considering the problem of
> image orientation.
> In DICOM files, the parameters of 0020 0037 indicate the directions of
> the first row and column. So we could use them to identify the
> original orientation.
> When we convert the DICOM files into NIFTI format, the converter add a
> transformation matrix. When we look at those images in different
> viewing softwares, they choose whether to adopt the transformation or
> just display the original images.
> When it comes to SPM8, it seems that SPM could display the images in
> classical views, but won't automatically change the directions for a
> unified final orientation. If the top-down direction is wrong,
> normalization couldn't be right performed.
>  
> I guess the NIFTI format can't solve everything? We still have to put
> the images into the right orientation before processing?
> So, if 0020 0037 in my dicom headers indicate roughly (1 0 0, 0 1 0) ,
> it means my images are in radiological orientation? and I have to flip
> them left-right to feed them into SPM? Am I right?
>  
> If I want to use reorientation tools in SPM to flip the images upside
> down and left to right, would a matrix [-1 0 0 0; 0 1 0 0，0 0 -1 0，0
> 0 0 1 ] be ok? 
>  
> Thanks 
> Yu
> 
>  
> 在 2010年3月3日 下午12:52，黄沛钰 <[log in to unmask]>写道：
>         Dear experts:
>          
>         I'm trying to reorient my images using the reorient utility. I
>         don't quite understand how the reorientation matrix works. If
>         I want to flip the images top-down, is [1 0 0 0; 0 1 0 0，0 0
>         -1 0，0 0 0 1 ] correct? Does it "only" flip the images
>         top-down? 
>          
>         Best
>         Yu 
> 


-- 
John Ashburner <[log in to unmask]>