On Tue, Jun 1, 2010 at 6:51 AM, Jonathan Peelle <[log in to unmask]> wrote:

Hi Graeme

> I thought I'd ask the collective wisdom and experience of the SPM list if
> they've ever seen out of brain clusters like those in the attached image.
> These are first-level individual subject SPMs at a FDR 0.05 10 voxel
> threshold. A number of subjects are showing highly significant clusters way
> outside the skull. Some are in the bone, but many are multiple mm away from
> any tissues, and I wonder what might be causing them? Could residual motion
> in anyway be responsible?

Having signal outside the brain isn't all that unusual, and in fact,
because the variance is low, these can even be quite significant.
However, generally these are masked out by the proportional
thresholding SPM does by default. So the first thing I would check is
the mask.img in the 1st level stats directory, which shows all the
voxels that SPM included in the analysis. Generally this looks
something like a brain mask, and if not, it would be good to figure
out why.

(Of course, your initial thoughts are good as well---checking that
realignment and normalization worked properly are the first things I
would check, I'm just assuming from what you've said you've checked
that already.)

> I'm concerned it is some kind of scanner noise, and it seems to be appearing
> more frequently in my recent scans, and need to decide if we should suspend
> scanning.

To evaluate potential scanner noise I would recommend some sort of
timeseries analysis that gives an indication of scan-to-scan variance.
I've had good luck with tsdiffana:

http://imaging.mrc-cbu.cam.ac.uk/imaging/DataDiagnostics

but there are several other options out there, which you may find by
digging into the list a bit. If you're worried about scanner noise it
may also be worth talking to physicists about phantom results, if
those are available... (e.g. Friedman & Glover (2006),
http://dx.doi.org/10.1002/jmri.20583).

Good luck!

Jonathan