I do peds research as well and some clinical scans, so I have used ArtRepair quite a bit.

1. The mask can be generated by "art_automask". It writes the mask to file "ArtifactMask.img". I have used this and it has always looked fine to me. I've never compared to the SPM generated mask, but I suspect it wouldn't be too different.
2. Yes, you should input the rp*.txt file(s).
3. For the pipeline, I think this will require some manual intervention on your part (i.e., it will depend on your data). You can see the doc here: http://cibsr.stanford.edu/content/dam/sm/cibsr/documents/tools/methods/artrepair-software/MotionandDespike.pdf
4. It's been awhile since I looked, but I don't remember there being a problem, in terms of motion parameters, when there is a large amount of between-session motion. However, if you look in the header of "art_global", he recommends doing the realignment and repair for each session separately. You may want to try different methods and see which makes the most sense.

You can alternatively (if you have a batch/script to go from raw images --> 1st level design estimation) run "art_global" and then "art_redo" afterwards. I don't know if this would be better or worse than inspecting the images at intermediate steps, so it might be worth the time to manually check your data instead of just automating everything.

Chris
________________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] on behalf of Sam DeWitt [[log in to unmask]]
Sent: Thursday, January 14, 2016 14:10
To: [log in to unmask]
Subject: [SPM] Using ArtRepair in preprocessing pipeline and multiple sessions

Hi,


I am analyzing data from a pediatric population with relatively high motion across 4 consecutively run sessions.  I'm attempting to use ArtRepair(version4) Bad Volumes: detect and repair option. This detects changes in the global signal and motion above set thresholds and allows them to be repaired using interpolation from nearest "good" frames(this is the option i'm using, rather than the de-spiking option which uses immediate frames before and after).   The GUI asks for a couple of prompts I had questions about and then I'm also seeing some conflicting reports on which pre-processing stage is best to do the repair, based on differing pipelines. Lastly, I wanted to get input on realignment with session vs. realignment to first session.

For the prompts using Bad volumes: detect and repair option in the GUI, the first prompt asks about which mask you want to use. Thus far I've just used the SPM generated mask. I can't seem to find info in the help option for this script as to how this mask is derived. Is it advisable to use, or better to create a user specific mask.  The second prompt asks to input the motion parameters. For this, so far I've been inputting the motion parameters text file that is generated from my realignment step of pre-processing, is this what is needed here?

I am attempting to incorporate this ArtRepair step into a standard pipeline in place. I've listed the pipeline steps below. Based on what I've read in the ArtRepair manuals it seems to make sense to do the repair after my realignment and slice time correction steps. That is what I've been doing, but am unsure  if this causes issues in the later processing steps.

1.      Segment T1 in native space(separate step from EPI images)
2.      Realign
3.      Slice time correction
4.      Spatially normalize to MNI
5.      Smoothing using 6mm kernel


Lastly,  standard practice has been to register the first frame of each session to the first frame of the first session and then process each session separately. As might be expected, this results in a lot of motion between sessions for this pediatric population based primarily on the realignment to that first sessions first frame(i.e. the session does not exhibit a lot of motion throughout, but the realignment of the first frame demonstrates gross head motion between sessions).  My inclination then is do all pre-processing steps within session(i.e. not register first frame of each session to first frame of first) but am unsure of potential problems this will cause at my first level analysis when I am concatenating sessions for my GLM.

Thank you in advance for help with some/all of my questions. I greatly appreciate it!

Best,
Sam DeWitt