Dear Caroline,
As far as I remember Artifact Detection Tools (ART) https://www.nitrc.org/projects/artifact_detect only detects outliers based on the rp files and the global signal, but it does not provide any options to write out corrected/filtered volumes, in contrast to ArtRepair toolbox http://cibsr.stanford.edu/tools/human-brain-project/artrepair-software.html (for which a beta SPM12 version has been released recently), which includes some quality check/corrections on the slice level (e.g. for spike artefacts) and provides options to correct bad data.
For ART I would rather not go with the rf* files, as due to the interpolation step from "Realign: Estimate & Reslice" it might become more difficult to detect intensity outliers originally present in the data. Also note that the reslice part is usually only necessary to obtain a mean EPI, but it is not necessary to also reslice the EPI series, as the realignment parameters are stored in the header and will be accounted for during the reslice part of "Normalise: Estimate & Write" or "Normalise: Write"), and the more reslice steps the more interpolation errors. Thus I would rely on the f* files.
It becomes somewhat more tricky if you use slice timing, as ST will lead to spreading of artefacts from one volume to the temporally neighbouring ones, but ST can also be expected to somewhat reduce the artefacts in the bad volumes (due to interpolation from neighbouring good ones). To stay on the safe side you could add dummy regressors for all those volumes pre-/succeeding a bad volume.
Best
Helmut
|