Hi Paul,
Upon comparing the motion parameters plots of my group, I think I have
2 patients that simply moved too much during the scan.
When I tried to 'repair' the slices, I often stumbled into the problem
that the slices in the n-1 and n+1 scans also were also not good, and
so the algorithm ended up repairing every slice. When I calculate how
many scans needed to be discarded (according the the parameters in
Lemieux, i.e. |d'| > 0.2 mm/scan & 12 seconds of nulling), then the
values are 136/286 and 247/286 scans. But then these subjects were
'jerking' continuously in the scanner. I'll still run the analyses
(with additional scan-nulling regressors), because my task should be
fairly robust, but still, especially the second scan should probably
not be analyzed.
Thanks for your reply in any case!
Regards,
Johan
On Tue, Oct 27, 2009 at 10:29 PM, Paul Mazaika <[log in to unmask]> wrote:
> The ArtRepair toolbox is a set of methods to help process fMRI data
> from high motion clinical subjects. It assumes the data has volumes with large
> slow motions that cause interpolation artifacts after realignment,
> and rapid motions that cause the distortions you mentioned (as well as
> errors in the realignment calculation).
>
> The pipeline preprocesses the data in an unusual way (HBM 2009 poster),
> namely, reslice after realignment, smooth by FWHM =4, then motion adjust and artifact repair,
> then normalize and smooth by FWHM =7 mm. The first smoothing makes
> the algorithm work better, the second one gives about the same
> smoothness as one FWHM =8 mm smoothing. These preprocessing steps "condition"
> the data to suppress large and rapid motion artifacts, and then
> the design matrix can be set up without any motion regressors.
>
> (These methods address the same motion and head jerk problems as in
> Lemieux 2007. However, that paper adds regressors to solve these issues,
> while the ArtRepair methods use more preprocessing.)
>
> The slice_repair function was written to filter out slice-wise electronic
> transients from a scanner. I don't know how it works for spin history effects.
>
> The repairs are "gentle", and try to only repair data that
> is excess variation beyond that expected by a BOLD effect. Thus, it
> makes a guess about how to divide up "task-correlated motion" effects
> between the cognitive task and excess that may be an artifact. Note that
> sometimes task-correlated motion can falsely inflate an activation
> (Hajnal, 1994), so the "more correct" answer for a BOLD estimate could be
> smaller. Also, task-correlated artifacts (e.g. from rapid motion) and
> deep breaths may also be a confound.
>
> These methods seem to work well on many high motion cases, but some
> subject data is still too difficult to analyze accurately. It is
> important to quality check the results for each subject
> (review the contrasts) and make sure they are reasonable.
>
> Good luck!
> Paul
>
> ----- Original Message -----
> From: "Johan" <[log in to unmask]>
> To: [log in to unmask]
> Cc: "Paul Mazaika" <[log in to unmask]>
> Sent: Tuesday, October 27, 2009 6:43:51 AM GMT -08:00 US/Canada Pacific
> Subject: Re: [SPM] Art_Repair, Art_Global
>
> Hello all,
>
> I am currently struggling with fMRI on some clinical patients who
> moved during the scan. Even though adding (a volterra expansion of)
> movement parameters seems to be widely recommended, I am loath to do
> so. The motion parameters have some correlation with the execution of
> the task, and including them therefore removes (too) much of the
> activity (motor cortex, cerebellum) if motion is severe. Something
> similar is hown in figure 2 of the article of Lund, 2005. (session 8).
>
> Re-scanning would probably not help, since these are clinical patients
> that move during the scan. I need something that I can do to fix
> motion artifacts from the EPI scans from clinical patients, without
> necessarily adding the motion parameters.
>
>
> In my understanding (from
> http://www.fil.ion.ucl.ac.uk/spm/toolbox/unwarp/), these are the
> motion-related contributions:
>
> 1) distortion by movement
> 2) total dropout by movement
> 3) spin-history effects
> 4) slice-to-volume effects
>
> Using the unwarping toolbox, the fist effect could perhaps be handled.
>
> Would the art_repair toolbox be able to deal with the slice-to-volume
> effects, by repairing bad slices?
>
> Would the scan-nulling approach (as in Lemieux, 2005) be accurately be
> able to deal with 3)?
>
> I am also wondering if there are any more motion-related effects apart
> from 1-4, and if anyone else has tried to deal with severe motion
> without including motion parameters in the design matrix.
>
> Does anyone have any experience with the mcica approach as outlined in
> Liao(2006)? There are a lot of approaches on dealing with motion. Are
> there other things I could try?
>
>
> Thanks in advance!
>
> Regards,
>
> Johan van der Meer
> PhD Stud.
> Academic medical centre, Amsterdam
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> On Wed, May 13, 2009 at 6:56 PM, Paul Mazaika <[log in to unmask]> wrote:
>> The art_global function tries to automatically detect bad image volumes according
>> to the criteria of unusual global signal, or large scan-to-scan motion,
>> or large total motion from baseline. Use it after realignment, because
>> it uses the rp.txt file. When used just before estimation, it is
>> easier to compare estimation results with and without it, to see if
>> it helps on a particular data set.
>>
>> The clip function marks a volume for repair if the total motion from
>> baseline is > 3mm, which may be useful for short duration artifacts when
>> no motion regressors are used in the design matrix.
>>
>> Handling artifacts is tricky, and the best method may not yet be known.
>> Consequently, the admittedly limited "manual" on the website does not give
>> strong guidance on the best choices of parameters.
>>
>> Note there are alternative (and better documented!) approaches to the artifact
>> problem, including removing volumes by inspection (Luo and Nichols, 2003)
>> or by the art_detect utility (Whitfield-Gabrieli), adding scan null regressors
>> (Lemieux, 2007), and adaptive weighting of the scans (RobustWLS toolbox).
>> Also, motion regressors (e.g. Lund, 2005) are helpful for removing
>> the effect of many artifacts.
>>
>> -Paul
>>
>>
>>
>> ----- Original Message -----
>> From: "Sanne Boesveldt" <[log in to unmask]>
>> To: [log in to unmask]
>> Sent: Wednesday, May 13, 2009 7:15:12 AM GMT -08:00 US/Canada Pacific
>> Subject: [SPM] Art_Repair, Art_Global
>>
>> Dear all
>>
>> I am trying to use art_repair/art_global in SPM5, but what exactly does
>> Art-global do? And when do I use it, after realignment, or after
>> normalization and smoothing? And what is the function of the 'clip' button
>> in the program?
>>
>> The only manual I could find are the short .txt and .pdf files from
>> http://cibsr.stanford.edu/tools/ArtRepair/ArtRepair.htm, but those don't
>> give that much info unfortunately. Is there a better manual out there?
>>
>> Thanks!
>>
>> Sanne Boesveldt, PhD
>> Postdoctoral Fellow
>> Monell Chemical Senses Center
>> 3500 Market Street
>> Philadelphia, PA 19104
>> USA
>>
>> +1 267 519 4688
>> [log in to unmask]
>>
>> --
>> Paul K. Mazaika, PhD.
>> Center for Interdisciplinary Brain Sciences Research
>> Stanford University School of Medicine
>> Office: (650)724-6646 Cell: (650)799-8319
>>
>> CONFIDENTIALITY NOTICE: Information contained in this message and any
>> attachments is intended only for the addressee(s). If you believe
>> that you have received this message in error, please notify the
>> sender immediately by return electronic mail, and please delete it
>> without further review, disclosure, or copying.
>>
>
> --
> Paul K. Mazaika, PhD.
> Center for Interdisciplinary Brain Sciences Research
> Stanford University School of Medicine
> Office: (650)724-6646 Cell: (650)799-8319
>
> CONFIDENTIALITY NOTICE: Information contained in this message and any
> attachments is intended only for the addressee(s). If you believe
> that you have received this message in error, please notify the
> sender immediately by return electronic mail, and please delete it
> without further review, disclosure, or copying.
>
|