There are a number of different approaches that you can use. If the images
are relatively well normalised already, then you shouldn't need to normalise
them again. One approach that we have tried recently to improve the
spatial normalisation for VBM is to base it entirely on segmented grey matter.
i.e., segment the raw images, spatially normalise by matching (without brain
masking) the grey matter segments to the apriori/grey.img. Then segment
these spatially normalised images, and strip off any remaining scalp.
Because the segmentation biases the results by overlaying the prior probability
images, then it is probably best to have these images overlayed systematically
so you have a similar bias for all subjects. By having the images spatially
normalised all the same way before segmenting, then the prior probability images
are overlayed the same for all subjects.
I hope this is helpful,
-John
| Thanks for this, it removed the scalp outline from the images very well.
| However, before I analyse the cleaned up segmented images, should I
| normalise them again?
|
| Many thanks
|
| Xavier
|
|
| >
| >| I am trying to analyse some old T1-weighted scans. They were done on a 1.0T
| >| scanner, and so the image contrast is not good. When normalising and
| >| segmenting the scans, a thin ribbon of scalp is consistently left around
| >| the brain. I have analysed plenty of other scans very successfully, so I
| >| think this is due to poor scan quality. However, before I give up, I was
| >| wondering if anyone had any suggestions.
| >
| >This may be due to lower scan quality, but often a thin ribbon of scalp is
| >to be expected. The segmentation does not yet work perfectly, and it does
| >not fully know the difference between brain and non-brain when the voxels
| >have similar intensities. However, it may be possible to automatically
| >remove most of this ribbon by the <Xtract brain> button (hidden in the
| >render pulldown).
| >
| >Select the xxx_seg1.img and xxx_seg2.img images that need to be cleaned up,
| >and specify the "Save Extracted Brain" option. This produces an image
| >called brain_xxx.img. Then you would use the <ImCalc> button to multiply
| >xxx_seg1.img by brain_xxx.img:
| >
| >Select xxx_seg1.img and brain_xxx.img, and specify an output filename.
| >The expression to be evaluated would be:
| > i1.*i2
|
| Xavier Chitnis
| Section of Cognitive Psychopharmacology
| Dept. of Psychological Medicine
| Institute of Psychiatry
| De Crespigny Park
| London, SE5 8AF.
| Tel: 0171 919 3977
| Pager: 07669 028 121
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|