Hi Richard
I don't know if you should trust me to "outline a consensus" for you,
but I can get you started on a couple of things.
One good script (surely there are others) is tsdiffana (time series
difference analysis), available from:
http://imaging.mrc-cbu.cam.ac.uk/imaging/DataDiagnostics
This give you a sense of the overall variability over time, and can
help you identify volumes with large variance (either image-wise, or
for some slices).
With respect to 'repairing' artifacts, I have no experience. However
I believe that several people have had good success modeling out bad
scans, an approach I think is mentioned here:
Lemieux et al. (2007) Modelling large motion events in fMRI studies of
patients with epilepsy. Magnetic Resonance Imaging 25:894-901.
http://dx.doi.org/10.1016/j.mri.2007.03.009
The idea being that adding an additional regressor for each bad scan
(where "bad" may be too much motion, or too much variance) will remove
the unique contribution of that (noisy) scan to your model.
Hope this helps,
Jonathan
On Tue, Apr 28, 2009 at 4:22 PM, Richard Binney
<[log in to unmask]> wrote:
> Hi all,
>
> I have a Rookie question for you expert users of SPM5 and SPM8 and 4D
> time-series Analyze fMRI data files. How do you assess your raw time series
> in terms of quality control?: by that I mean inspection for volume and
> slice artifacts, large movement and spike artifacts. And what about the
> subsequent repair?
>
> Not once have I seen an SPM or fMRI analysis lecture which mentions
> inspection and subsequent repair of the raw time-series and the appropriate
> way to do this given the file format we use. Is this because it is assumed
> that the way we analyze the data using SPM largely accounts for these
> artifacts in some way, and therefore we don't need to do it?
>
> Quality inspection has been mentioned on the list: Unlike programs like
> AFNI, SPM5 (or mricro) doesn't seem to have an intuitive means to inspect
> time series. Of course you can plot voxel, ROI or cluster time series but I
> can't see a way of displaying an slice image in the time dimension like in
> anlyze movie or AFNI. I have noticed that some people use 'ArtRepair'
> and/or 'Anlayzemovie' to check the time-series and sometimes correct it.
> Are these widely used utilities/toolboxes? Are there other ways which I can
> display the time series as a image or as a plot?
>
> If someone could outline a consensus for me, it would be very useful. for
> future reference.
>
> Your gratefuly,
>
> richard
|