And can I put in a good word for Joern's approach, which works how how
noisy each image is and then gives less weight to the bad ones. You
don't have to classify an image as 'good' or 'bad', the algorithm
gives it a weight instead. It works for my data.
details here:
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=pubmed&dopt=Abstract&list_uids=15975828&query_hl=19
Antonia.
On 8/3/05, Darren Gitelman <[log in to unmask]> wrote:
> Dear Bartosz, Steve, Russ and SPM:
>
> My additional 2 cents: One other type of motion to keep in mind is
> interslice or within volume motion. This is particularly apparent if using
> an interleaved sequence and the subject moves somewhere in the middle of
> the acquisition- the volume then looks like an incompletely shuffled deck
> of cards. None of the algorithms in common use can correct for this. We've
> found that Matthew Brett's tool tsdiffana is very good at picking up this
> type of motion.
>
> Darren
>
> At 12:25 PM 8/2/2005, Russ Poldrack wrote:
> >well - they are sort of the same thing, since the resliced images are
> >interpolated using the estimated motion parameters (which I think is what
> >is presented in the graph). It's important to keep in mind that these
> >parameters are based on the assumption that motion effects are rigid-body;
> >if motion has more complex effects (e.g., interaction with susceptibilty,
> >striping in an interleaved sequence) they will not be accurately estimated
> >or corrected using this method.
> >
> >cheers
> >russ
> >
> >On Aug 2, 2005, at 10:13 AM, Steve Cramer wrote:
> >
> >>Does anybody know: are the graphs that SPM produces with motion
> >>correction graphs of motion detected, or of motion corrected?
> >>--Steve
> >>
> >>
> >>
> >>
> >>>Hello Lila,
> >>>two relevant things beyond what Russel mentioned is that rotational
> >>>motion is much more difficult to correct for and thus more critical when
> >>>deciding to further analyze the data or not. The 2nd issue is whether
> >>>your motion peaks are correlated with your events or not. If so, motion
> >>>correction might strongly influence your signal.
> >>>Regards, Bartosz
> >>>
> >>>
> >>>>On Aug 2, 2005, at 7:54 AM, Lila Davachi wrote:
> >>>>
> >>>>
> >>>>>Hi all
> >>>>>I am looking for some guidance into how to decide if a subject has
> >>>>>simply moved "too much", in other words, how much motion is too much
> >>>>>motion?
> >>>>>
> >>>>>I have been using a shorthand guide and only have included Ss who
> >>>>>moved less then one voxel (in this case, 3mm) and trusted that
> >>>>>motion correction could handle that. But does the kind of motion
> >>>>>matter? For example, I have a subject now who only moved ~ 2mm across
> >>>>>a 1/2 hour scan but the motion was concentrated at specific times
> >>>>>with sharp peaks in the graph.
> >>>>>
> >>>>>So, my question is how can I convince myself that SPM has done a good
> >>>>>job at correcting the motion?
> >>>>>What are some tests of the data that I can implement (besides
> >>>>>including motion as a regressor..).???
> >>>>>
> >>>>>thanks much
> >>>>>lila
> >>>>>
> >>>>
> >>>>---
> >>>>Russell A. Poldrack, Ph.d.
> >>>>Assistant Professor
> >>>>UCLA Department of Psychology
> >>>>Franz Hall, Box 951563
> >>>>Los Angeles, CA 90095-1563
> >>>>
> >>>>phone: 310-794-1224
> >>>>fax: 310-206-5895
> >>>>email: <mailto:[log in to unmask]>[log in to unmask]
> >>>>web: <http://www.poldracklab.org>www.poldracklab.org
> >>>>
> >>>>
> >>>
> >>>--
> >>>Dr. Bartosz Zurowski
> >>>Klinik für Psychiatrie und Psychotherapie
> >>>UKSH Campus Lübeck
> >>>&
> >>>Department of Systems Neuroscience
> >>>NeuroImage Nord, Bldg. S10, UKE
> >>>Martinistrasse 52 phone: 040-42803-3683
> >>>20246 Hamburg fax: 040-42803-9955
> >>>
> >>
> >
> >---
> >
> >Russell A. Poldrack, Ph.d.
> >
> >Assistant Professor
> >
> >UCLA Department of Psychology
> >
> >Franz Hall, Box 951563
> >
> >Los Angeles, CA 90095-1563
> >
> >
> >phone: 310-794-1224
> >
> >fax: 310-206-5895
> >
> >email: <mailto:[log in to unmask]>[log in to unmask]
> >
> >web: <http://www.poldracklab.org>www.poldracklab.org
>
|
