I think that Marko is correct in that it is worth investigating the image inhomogeneity and bias correction parameters. If the original alignment is ok, as you say, you may be able to try different bias correction settings. Alternatively, you can do a two-pass bias correction procedure, which is probably less elegant but often seems to work. To do this, run segmentation once and output the bias-corrected image, but no segmentations. Then run segmentation again on this bias-corrected image. For images with significant inhomogeneity this often works wonders.
On Feb 23, 2011, at 10:17 AM, Marko Wilke wrote:
> Hello Justine,
> this looks like the initial affine registration did not work, perhaps due to image inhomogeneity. Aligning the images nicely, for example by coregistering to a template prior to segmentation, may help, as may fiddling with the inhomogeneity parameters (bias FHWM etc.).
> Good luck,
> justine dupont wrote:
>> Dear experts
>> I recently saw something very puzzling with a VBM analysis. I have 40 T1
>> scans. They all have the same dimensions and voxel size and the raw
>> images all look consistent and normal. Segmentation (with normalisation
>> and modulation) works fine for 38 but for one the GM result is an empty
>> brain (top row of attached image) and for the second the result is a
>> kind of 'half brain' (bottom row of attached image).
>> What causes this and can it be fixed?
>> Many thanks
> PD Dr. med. Marko Wilke
> Facharzt für Kinder- und Jugendmedizin
> Leiter, Experimentelle Pädiatrische Neurobildgebung
> Abt. III (Neuropädiatrie)
> Marko Wilke, MD, PhD
> Head, Experimental Pediatric Neuroimaging
> University Children's Hospital
> Dept. III (Pediatric Neurology)
> Hoppe-Seyler-Str. 1
> D - 72076 Tübingen, Germany
> Tel. +49 7071 29-83416
> Fax +49 7071 29-5473
> [log in to unmask]