Thank you for the suggestion Volkmar. It is possible - I had done a few co-registration steps on the images with the aim of getting them in the same space (they are from different scans) so something may have gone wrong.
When I check in SPM>Display Image,
the 2 images have the same voxel size but differ slightly in the Origin
64.2 69.7 19.9 vs. 63.5 72.7 16.6
one is uint8 and the other uint16, and the ‘Direction cosines’ values also differ slightly
these 2 particular images produce 61 voxels when convolved with ImCalc, and 30 voxels when convolved with my own script
(image 1 has 143 voxels, and image 2, 1714)
So it seems likely that the problem is related to a difference between the 2 images which is only taken into account in the spm computation as you suggest. Given that the difference in e.g. the origin between the 2 images is so small, do you (or anyone else) think my coregistraton steps were still correct, or have I probably done something wrong? Basically I want to know if it’s possible to get absolutely perfect coregistration between images from different scans, or whether this small difference between the images is to be expected?
thanks,
Susan
On 19 Sep 2014, at 11:29 pm, Volkmar Glauche <[log in to unmask]> wrote:
> Dear Susan,
>
> would it be possible that your images have the same matrix dimensions, but different orientation or voxel size? spm_imcalc will take that into account by reslicing the second image into the space of the first, while your script does not. Different orientation would explain why the computed overlap between the two masks would differ between the two approaches.
>
> Best,
>
> Volkmar
|