Hi Lukasz,
> 1. I have incomplete head acquisitions so that I am looking for a way to make FIRST run on a restricted range of standardized MNI Z-coordinates in a similar way that is possible with SIENAX (for instance, -t 60 -b -50). I would also like to have in my outputs head images that are cut at those standardized coordinate values by in native space.
FIRST’s segmentation algorithm uses only local image information, so the only thing it requires is that each of the supported structures (plus a small margin) is in the field of view. From looking at the MNI template, for the slab you mention, I think this would mean that you would run into problems with the brainstem, but the other structures should work fine.
The default registration procedure (the first_flirt script that is invoked by run_first_all) does expect to see more-or-less whole brain scans, but I would expect it to work fine with the slab you mention.
Are there any specific problems that you are running into? Note that SIENAX needs this information because summing over an incomplete brain would change the tissue volume estimates. In FIRST this is not an issue unless you have incomplete coverage of the subcortical structures of interest.
> 2. Are volume measurements generated by FSL-FIRST normalised for head size such as in SIENAX, and if not, what normalisation process would be suitable (e.g. multiplication by the scaling factor derived from SIEANX?, other suggestions?)
The fslstats command that is mentioned in the manual computes the size of the structure in a subject's original (native) scan, so this is not corrected for brain size. We normally recommend using the VSCALING factor from SIENAX.
> 3. I would like to compare FSL-FIRST measurements with those performed with Lesion Toads (part of MIPAV). For this, I would need to have head mages that are cut at the same MNI Z-coordinates as specified earlier in FSL-FIRST (e.g. -t 60, -b -50 when used SIENAX notation).
If you want to mask your images using a slab based on MNI coordinates, one way of doing this would be to label your slab on the MNI template. E.g.:
fslmaths $FSLDIR/data/standard/MNI152_T1_1mm -mul 0 -add 1 -roi 0 -1 0 -1 22 110 0 -1 roi
You can then register this ROI to each of your scans. Note that the parameters to -roi are in voxels, not mm.
> 4. If I do everything successfully, I still have a question regarding the registration used for Lesion-Toads measurements. When comparing brain volume measurements carried out with FSL-SIENAX, the authors of Lesion-Toads registered images rigidly (6 DOF) to a standard brain and then performed measurements with Leison-Toads ("we rigidly registered the stripped MPRAGE scans to a single image atlas (“JHU_MNI_SS_T1”) " in http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0037049). My question is whether this rigid registration could interfere with the volumetric outputs as compared to native images (i.e. will volume measurments of native brains and such registered brains be different).
It seems to me that rigid body registration shouldn’t change the volumes, but it is probably best to ask the Lesion-Toads authors.
Hope that helps,
Eelke
|