Skip to content

Latest commit

 

History

History
95 lines (67 loc) · 3.66 KB

Process_DIAGNOSE-CTE_data.md

File metadata and controls

95 lines (67 loc) · 3.66 KB

You must read Process_HCP-EP_data.md before starting off with this tutorial.

Recap

Execution of all Luigi tasks require three things:

  1. Source the proper environment e.g. bashrc3, HD-BET, CNN-Diffusion-MRIBrain-Segmentation.
  2. Define the proper configuration export LUIGI_CONFIG_PATH=/path/to/config.cfg
  3. Formulate the /data/pnl/soft/pnlpipe3/luigi-pnlpipe/exec/ExecuteTask command

Structural pipeline

Its description is the same as that of HCP-EP structural pipeline. Instead of HD-BET, we used our good old MABS (Multi Atlas Brain Segmentation) algorithm to create T2w masks.

It is run in two steps: T2w mask creation and FreeSurfer segmentation.

  • T2w mask creation
source /data/pnl/soft/pnlpipe3/bashrc3
export LUIGI_CONFIG_PATH=/data/pnl/soft/pnlpipe3/luigi-pnlpipe/params/cte/T2w_mask_params.cfg
/data/pnl/soft/pnlpipe3/luigi-pnlpipe/exec/ExecuteTask --task StructMask \
--bids-data-dir /data/pnl/DIAGNOSE_CTE_U01/rawdata \
--t2-template "sub-*/ses-01/anat/*_T2w.nii.gz" \
-c 1004 -s 01
  • FreesSurfer segmentation
source /data/pnl/soft/pnlpipe3/bashrc3
export LUIGI_CONFIG_PATH=/data/pnl/soft/pnlpipe3/luigi-pnlpipe/params/cte/struct_pipe_params.cfg
/data/pnl/soft/pnlpipe3/luigi-pnlpipe/exec/ExecuteTask --task Freesurfer \
--bids-data-dir /data/pnl/DIAGNOSE_CTE_U01/rawdata \
--t1-template "sub-*/ses-01/anat/*_T1w.nii.gz" \
--t2-template "sub-*/ses-01/anat/*_T2w.nii.gz" \
-c 1004 -s 01

Diffusion pipeline

It is run in three steps--T2w mask creation, DWI mask creation, and FSL eddy correction:

  • Step-1

The T2w mask is created as part of structual pipeline. This mask is warped to the space of an axial-T2 image. Hence, there is a line in the above flowchart going from StructMask (T2w) node on the left to StructMask (AXT2) node on the right. The axial-T2 image is used for EPI correction.

  • Step-2

DWI mask is created using our own CNN-Diffusion-MRIBrain-Segmentation tool. It is a deep learning based brain extraction tool. It should be run on a GPU device i.e. grx** node or bhosts gpu_hg cluster. To run the DWI masking program through bhosts gpu_hg cluster, you may use run_gpu_mask.lsf as an example.

source /data/pnl/soft/pnlpipe3/CNN-Diffusion-MRIBrain-Segmentation/train_env
export LUIGI_CONFIG_PATH=/data/pnl/soft/pnlpipe3/luigi-pnlpipe/params/cte/cnn_dwi_mask_params.cfg
/data/pnl/soft/pnlpipe3/luigi-pnlpipe/exec/ExecuteTask --task CnnMask \
--bids-data-dir /data/pnl/DIAGNOSE_CTE_U01/rawdata \
--dwi-template "sub-*/ses-01/dwi/*_dwi.nii.gz" \
-c 1004 -s 01

After checking quality of the automated mask, it must be saved with Qc suffix in the desc field for its integration with later part of the diffusion pipeline. Example:

Automated mask  : sub-1004/ses-01/dwi/sub-1004_ses-1_desc-dwiXcUnCNN_mask.nii.gz
Quality checked : sub-1004/ses-01/dwi/sub-1004_ses-1_desc-dwiXcUnCNNQc_mask.nii.gz

Pay special attention to the string CNNQc_mask.nii.gz

  • Step-3

Finally, run EDDY correction followed by EPI correction:

source /data/pnl/soft/pnlpipe3/bashrc3
export LUIGI_CONFIG_PATH=/data/pnl/soft/pnlpipe3/luigi-pnlpipe/params/cte/dwi_pipe_params.cfg
/data/pnl/soft/pnlpipe3/luigi-pnlpipe/exec/ExecuteTask --task EddyEpi \
--bids-data-dir /data/pnl/DIAGNOSE_CTE_U01/rawdata \
--dwi-template "sub-*/ses-01/dwi/*_dwi.nii.gz" \
--t2-template "sub-*/ses-01/anat/*_AXT2.nii.gz" \
-c 1004 -s 01