Analysis code and Flywheel gear files for TOME fixel-based diffusion analysis
- Python3 (Python2 should also work but not tested)
- ANTs
- AFNI (If the adaptive convex method will be used)
- Mrtrix
- trekker https://dmritrekker.github.io/manual/trekker.html
- FSL
All analysis functions are wrappers around one or multiple software mentioned above. All of the steps mentioned next were run with Flywheel gears. The scripts to submit these gears for all subjects can be found in submitGears folder of this repo.
1- Creating a FOD image from HCP-diff results.
- For adaptiveConvexOptimization Tran&Shi(2015), fodMakerWrapper.m could be run directly with the DTI images and bval/bvec files.
- For the mrtrix method, first run calculateAverageResponseFunction.py to calculate 3-tissue average response functions and then run create_fod_image.py to make FOD images. This method was used for the TOME data. We used multi-shell option since our data had multiple b-values with 8 lmax.
2- We run create_fod_population_template.py to calculate an unbiased FOD template image.
3- An anatomical template was created with ANTs from all subject MPRAGE images with the following call "antsMultivariateTemplateConstruction2.sh -d 3 -c 2 -A 1 -a 1 -b 0 -e 1 -g 0.2 -i 4 -j 6 -k 1 -w 1 -q 500x500x100x50 -f 6x4x2x1 -s 3x2x1x0 -n 1 -o /flywheel/v0/output/TOME -r 1 -l 1 -m CC -t SyN -y 1"
4- recon-all was run on the template.
5- The thalamic regions of the template was segmented following the Freesurfer instructions: https://freesurfer.net/fswiki/ThalamicNuclei
6- LGNs were extracted from the ThalamicSegmentation results with the Freesurfer mri_extract_label function using the segmentations codes 8109 and 8209 (left and right LGN respectively). The optic chiasm was extracted with the code 85
7- Extracted segmentations were linearly (12DOF) registered to the FOD template.
8- A wrapper around trekker (roi_tractography.py) was used to calculate the optic tract using the LGN and optic chiasm segmentations from the previous step.
9- fixelAnalysis.py was used to calculate whole brain fixel metrics.
10- With the extract_fixel_values.py function the whole brain fixel maps were cropped using the tracts from step 8. The function also produces optic tract fixel stats.
1- FSL's dtifit was used to fit tensors to the preprocessed DTI images.
2- extract_dti_values.py was used to construct whole brain DTI metric values as well as the ones cropped with the tracks calculated at the fixel analysis step 8.