Hi,

I hope you are doing well. I have processed approximately 100 datasets through the ARTS toolbox and just finished visual QC of the outputs. All look great except for a small subset, for which the WMH is substantially undersegmented due to registration issues between the T1w and FLAIR images. 

To solve this issue, is there a way to fix the co-registration and feed this into ARTS to regenerate the WMH map? Or would you suggest generating a binary WMH segmentation map via a secondary program (e.g., LST in SPM) and incorporating this map into a rerun of ARTS? 

Thanks in advance,
Jenna

Hello Jenna,

For participants with poor T1w-FLAIR co-registration, I suggest providing ARTS with a registration matrix from a successful registration. ARTS accepts the registration matrix generated by FSL’s FLIRT.

The following command is how flirt is executed in ARTS. Note that the command is executed in the participant’s folder. The input image must be FLAIR_n4_reorient.nii.gz, and the reference image must be T1_n4_reorient.nii.gz. Both images are generated by ARTS and can be found in the participant’s FLAIR and T1 folders. You can modify the registration options as you want, but I recommend keeping dof as 6 for rigid body registration.

flirt -in FLAIR/FLAIR_n4_reorient.nii.gz \

  -ref T1/T1_n4_reorient.nii \

  -searchrx -30 30 \

  -searchry -30 30 \

  -searchrz -30 30 \

  -usesqform \

  -coarsesearch 10 \

  -finesearch 5 \

  -dof 6 \

  -omat FLAIR/coreg_to_t1.mat \

  -inweight FLAIR/pre_brain_mask.nii.gz \

  -refweight T1/brain_mask.nii.gz \

  -out T1/FLAIR.nii.gz

Once you have an acceptable registration, copy the output matrix defined by the -omat option into a new folder called user_input. The user_input folder needs to be in the participant’s folder, at the same level as the QC folder, T1 folder, etc. The filename for the registration matrix must be coreg_to_t1.mat

Then, you can run ARTS on the participant again. ARTS will reprocess the data and use the registration matrix you provided instead of calculating the co-registration.

You can create a registration matrix and then rerun ARTS for each participant one at a time, or you can create the registration matrix for all the problematic participants and then rerun ARTS on all of them in one run. Let me know if anything is unclear in the above process.

-Arnold