Dear Jim,
I hope the ArtRepair developers or users will be able to answer your
questions but concerning your issue with 4D files, it should in theory
be possible to provide the input images as a single 4D filename "4D.nii"
or a list of filenames with explicit frames, eg "4D.nii,1" "4D.nii,2"
etc, without having to create 3D files, see:
https://en.wikibooks.org/wiki/SPM/Working_with_4D_data
A quick look at the code in ArtRepair seemed to imply this would work
but it would actually still require some tweaks. Given this, I would
convert the 4D dataset to a list of 3D files and proceed with these
throughout (ArtRepair and SPM).
Best regards,
Guillaume.
On 27/07/18 14:35, Jim Bjork wrote:
> Hello Folks,
>
> Having done my postdoc at NIH intramural, I've been using AFNI exclusively. However, in light of all the data I'm poised to lose due to motion in a study of Borderline Personality Disorder subjects, I'm trying to explore ArtRepair, as its approach seems very sensible, and my lab group does have and use SPM12. The only step-by-step guide to using the software appears to be this single web page http://cibsr.stanford.edu/tools/human-brain-project/artrepair-software/artrepairinstructions.html and there is no static-cam video of a narrating user going thru the GUI steps of an analysis like the nice SPM general series by Andrew Jahn on YouTube.
>
> Thus, I'm hoping Dr. Mazaika or another could confirm (or refute) the following understandings:
>
> 1) art_slice is not really going to help me unless there is scanner-driven artifact, such as noise spikes, especially those that might only affect one slice/location. I should focus on art_global.
>
> 2) unlike modules like AFNI 3dVolReg, there are no ArtRepair programs that can invoke and process a 4D (3d+time-series) .nii dataset directly. For either art_slice, art_global, or even art_movie, one must use "spm_file_split *.nii" to bust up the .nii file into sequentially-numbered individual volumes, invoke ALL volumes as individual files in the art-repair GUI, run the program, and use something like AFNI to3d to reconstitute the 3d+time file for further processing.
>
> 3) If I want to try to use art_global per 2) above, I should run it on a 3d+time volume (busted up into slices (?))that has been volume-registered first, such as with AFNI 3dDeconvolve, and somehow convert the .1D file that afni generates of the motion corrections over time into some kind of .txt file that SPM will work with, in order to get best results
>
> 4) the default threshold value (data field entry) represents the percent difference from mean intensity value, at which to flag and correct a volume with a nearest-temporal-neighbor interpolation, and the default values are generally reasonable.
>
> 5) If I can end up with a 3d+time dataset with select volumes replaced with interpolated volumes by art_global, where the motion text file was used in the interpolation, I now no longer need to include motion regressors in my task modeling.
>
> I would appreciate clarifications on these points and any other tips, if anyone has a moment.
>
> Thanks,
>
> Jim
>
--
Guillaume Flandin, PhD
Wellcome Centre for Human Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|