>>On Fri, 11 Jun 1999, John Ashburner wrote:
>> SPM'96 could create normalized images of whatever dimensions you asked.
>> By default however, it produced images with a bounding box of:
>> -78:78mm in the x direction, -112:76mm in the y direction, and -50:85mm
>> in the z direction. The default voxel size was an isotropic 2mm.
>> If the spatially normalized image trims off bits of the brain that you
>> are interested in, then you need to extend the bounding box.
>> if you are just interested in a limited region of the brain, then you can
>> specify a bounding box that just covers that region.
>On Fri, 11 Jun 1999, Vitaly Furman wrote:
>I must admit I was not clear enough when I posted the question. I always
>had the impression that the template used by SPM'95 (PET image with
>deleted skull called spm.img) corresponded to the bounding box used by
>SPM'95. The Maximum Intensity Projections printed out by SPM'95, had
>numbers that delineated the boundaries of the bounding box.
>SPM'96 does not print any numbers at the boundaries of the box. But the
>Maximum Intensity Projections look very similar to the ones in SPM'95 -
>the box does not extend beyond the contour of the brain. This makes sense
>since the Talairach atlas should be the same regardless of which bounding
>box was used. The grid that is printed on the MIPs, corresponds to the
>grid of standard Talairach atlas with dimensions (-64;64), (-104;68),
>(-28:72). So my question is: do the MIPs retain these dimensions even
>if a differentnbounding box was used at spatial normalization?
The bounding box is a separate issue from both the Talairach grid and the
template. The bounding box is ONLY used to crop the final output (spatially
normalised) images. Hence the dimensions of the brain in the MIPs remains
the same regardless of bounding box. A, possible, reason for using e.g. a
smaller bounding box is to save disc space in the event that one is
exclusively interested in looking at a specific part of the brain.
The dimensions of the brain is exclusively determined by the template, and
the calculations of the parameters for the spatial normalisation uses all
available data regardless of which bounding box has been specified.
The display (the MIPs) is also independent of the bounding box, and will
look the same regardless of what BB has been specified, although the will
of course be zero outside the BB.
I hope I have understood your question correctly, and that this clarifies
Good luck Jesper