This tutorial explains how you can integrate DTI-TK into the standard TBSS pipeline. This affords you the benefit of leveraging the state-of-the-art spatial normalization provided by DTI-TK together with the state-of-the-art voxel-wise statistical inference for white matter anatomy. For people who need an explanation of what TBSS is, please refer to this FSL page.
Update (August 2014): A recent NeuroImage article entitled Methodological considerations on tract-based spatial statistics (TBSS) recommends the use of DTI-TK to enhance TBSS.
As outlined in the TBSS documentation, there are altogether six components to the full analysis: 0) Preprocessing, 1) tbss_1_preproc, 2) tbss_2_reg, 3) tbss_3_postreg, 4) tbss_4_prestats, and 5) stats.
If you have successfully run DTI-TK to spatially normalize of your data into the corresponding group-specific DTI template, you have completed the first three components in the standard TBSS pipeline.
The integration of DTI-TK output into the TBSS pipeline requires a custom implementation of tbss_3_postreg, the goal of which is to generate the spatially normalized high-resolution DTI data, i.e., with a spatial resolution of isotropic 1mm3, the mean FA skeleton, and the 4-D FA map of the spatially normalized high-resolution DTI data. (:showhide init=hide div=tbss_3_postreg:)
dti_warp_to_template_group subjs.txt mean_final.nii.gz 1 1 1
subjs.txt is an ASCII text file that contains a list of the file names of the subject DTI volumes in their native space (after the DTI-TK pre-processing steps) and mean_final.nii.gz is the final population-specific DTI template for these subjects computed with DTI-TK. The last 3 arguments, not surprisingly, specify the desired voxel spacings. This command should be executed while your current working directory is the directory where you have run all your DTI-TK registration steps.
TVMean -in subjs_normalized.txt -out mean_final_high_res.nii.gz
subjs_normalized.txt is an ASCII text file that contains a list of the file names of the normalized high-resolution DTI volumes from the previous step. mean_final_high_res.nii.gz will be the output high-resolution DTI template.
TVtool -in mean_final_high_res.nii.gz -fa
mv mean_final_high_res_fa.nii.gz mean_FA.nii.gz
mean_FA.nii.gz is now the new name for the high-resolution FA map of the DTI template.
tbss_skeleton -i mean_FA -o mean_FA_skeleton
mean_FA_skeleton will be the white matter skeleton for running the subsequent TBSS analysis. The suffix (hdr/img or nii or nii.gz) depends on your FSL output configuration.
fslmerge to combine them into a 4-D volume called all_FA. Finally, apply fslmaths to all_FA to create a combined binary mask volume called mean_FA_mask. The command to do this
fslmaths all_FA -max 0 -Tmin -bin mean_FA_mask -odt char
tbss with a subdirectory called stats. Copy mean_FA_skeleton, all_FA, and mean_FA_mask to the stats subdirectory.
After completing this custom tbss_3_postreg, simply change to the directory of tbss, which you just created, and continue with the final two steps in TBSS.