Hi-
Re. #1: you can include that anatomical-space dataset as a follower, and it will be regridded to the final EPI space in the end. You can use one of hte following 2 options, depending on whether
- that dataset is integer-valued and should be interpolated with nearest-neighbor (NN) mode, to preserve integers; in that case, use:
-anat_follower_ROI LABEL GRID DSET : specify anat follower ROI dataset
e.g. -anat_follower_ROI aaseg anat aparc.a2009s+aseg_REN_all.nii.gz
e.g. -anat_follower_ROI FSvent epi fs_ap_latvent.nii.gz
Use this option to pass any anatomical follower dataset. Such a
dataset is warped by any transformations that take the original
anat to anat_final.
Similar to -anat_follower, except that these anatomical follower
datasets are resampled using nearest neighbor (NN) interpolation,
to preserve data values (as opposed to -anat_follower, which uses
wsinc5). That is the only difference between these options.
LABEL : to name and refer to this dataset
GRID : which grid should this be sampled on, anat or epi?
DSET : name of input dataset, changed to copy_af_LABEL
Labels defined via this option may be used in -regress_ROI or _PC.
- or it is floating-point valued, and can be interpolated with a cubic or something, in which case use:
-anat_follower LABEL GRID DSET : specify anat follower dataset
e.g. -anat_follower GM anat FS_GM_MASK.nii
Use this option to pass any anatomical follower dataset. Such a
dataset is warped by any transformations that take the original
anat to anat_final.
Anatomical follower datasets are resampled using wsinc5. The only
difference with -anat_follower_ROI is that such ROI datasets are
resampled using nearest neighbor interpolation.
LABEL : to name and refer to this dataset
GRID : which grid should this be sampled on, anat or epi?
DSET : name of input dataset, changed to copy_af_LABEL
A default anatomical follower (in the case of skull stripping) is
the original anat. That is to get a warped version that still has
a skull, for quality control.
From your description, it sounds like you have a map of GM, which would likely be integer valued, and so I guess using the first would be appropriate, and getting it to be on the EPI grid at the end could be done with, like:
-anat_follower_ROI seg_gm epi <dset_with_gm_mask>
Re. #2: I think this is made moot by the answer to #1. The EPI-anatomical alignment direction can be flipped, but it is just an intermediate step, and tends to work out better the way it is done in practice, and since the "follower" option gives you what you want for including the GM, then why mess with well-working alignment?
Re. #3: Oh, I see, you haven't included a "regress" block, actually. I misread that initially. Even if you don't care about the regression results, just add it at the end of your list of blocks, so:
-blocks align volreg mask regress
... in order to have the QC generated (otherwise it won't be).
And actually, since you are doing full resting state processing, you will want the regress block there. I would then also add in these opts for processing (you can choose what censor thresholds you want, these are kind of general starting points, or you can remove them if you want, but that seems risky with resting state):
-regress_apply_mot_types demean deriv \
-regress_motion_per_run \
-regress_censor_motion 0.2 \
-regress_censor_outliers 0.05 \
-regress_est_blur_epits \
-regress_est_blur_errts \
Some further blocks to consider:
- And you can scale your data with local voxelwise scaling, so the units have meaningful interpretations as local BOLD % signal change, which we would recommend over grand mean scaling (see here), just by adding a "scale" block to your list of blocks.
- While adding blocks for resting state, you might want to add a "despike" at the beginning, to remove some likely motion effects.
- if you have timeshift information, you could add a "tshift" block
- If you are doing voxelwise analysis, it would be common to blur the data a bit, so you can add a "blur" block and blur value, which for single-echo FMRI we would recommend something like 1.5-2 times your EPI's minimum voxel dimension.
So then your block list (and blur opt) could be:
-blocks despike tshift align tlrc volreg mask blur scale regress \
-blur_size BLUR_SIZE \
Again, much of this depends on your analysis goals.
--pt
