12.8cm - those are pretty thick slices. 12.8mm are also pretty thick, which
is what I assume you mean. There are a couple of issues here...
1) There is likely to be a lot of partial volume effect, which the SPM99
segmentation algorithm can not cope with particularly well. It assumes that
voxels only contain one tissue type, so doesn't work so well when they
contain a mixture of tissues.
2) The results will depend on inter-slice gaps. If there are large gaps
between the slices, then it is likely that you are acquiring completely
different parts of the brain. Unless all your scans are positioned in
exactly the same way, there probably isn't a way of getting reproducible
segmentations.
You could get an intuitive idea of how different images are by doing a
coregistration of the datasets, and looking at the registered images with
check-reg. No-matter what reslicing algorithm is used, there is no way of
recovering the missing information.
Best regards,
-John
> Another question regarding segmentation: when we segment sequentially
> acquired images from the same subject (that unfortunately have been
> acquired using only 12.8 cm slice thickness) it may happen that the
> reproducibility of the resulting gm, wm and csf fractions is poor. Can poor
> repositioning with respect to slice angle (like ‘pitching’) be the cause
> for this, as it obviously changes the amount of head included, ie the
> completeness of the lower slices available to analysis using the prior
> probability images? If so, would it be useful to reslice the images so that
> for a given subject the angle and thickness become the same (even if this,
> by cutting away lower parts, causes loss of pixels containing image
> information)?
> Hope I made myself clear, unfortunately example images may not be posted.
--
Dr John Ashburner.
Functional Imaging Lab., 12 Queen Square, London WC1N 3BG, UK.
tel: +44 (0)20 78337491 or +44 (0)20 78373611 x4381
fax: +44 (0)20 78131420 http://www.fil.ion.ucl.ac.uk/~john
|