DTI-TK
Last updated
Last updated
DTI-TK stands for Diffusion Tensor Imaging Toolkit. The main way we use DTI-TK in our lab is to register all diffusion images in a dataset to a common space. We use the registration portion of DTI-TK and integrate it with TBSS, using DTI-TK's own guidelines for doing so.
Generally, this process is done in five major steps we have collected in five scripts.
Inputs should be preprocessed (artifact corrected) DWI data, such as those output by PyDesigner:
The nifti input should be skull stripped. It is extremely important to check the skull stripping outputs as a bad skull strip will negatively affect registration downstream. During this process, a QC folder is created so that skull stripping outputs are easy to QC. If skull stripping is not sufficient, the skull stripping threshold (-f # in the script) can be modified. Depending on the dataset, BET can be highly customized to achieve appropriate skull stripping.
Due to the nature of tensor files and default scales in many image viewers, just looking at the skull stripped tensor with default scale settings may be deceptive:
The above image appears appropriately skull stripped but by overlaying the mask onto it, we see that it is not skull stripped correctly:
Once you refine the BET process, the mask overlay should look more like this:
After skull stripping is done, FSL's dtifit is used to create a tensor.nii file using a nifti and its associated bvec, bval, and mask (created using skull stripping). Then fsl_to_dtitk is used to convert FSL outputs to DTI-TK format.
Now we are ready to register our images.
The actual registration process is simple, but can be time consuming. Simply run each line in order.