Dear FSL experts,
first of all, thanks for the developing of such a useful software, for the continuous updating of functions, and for the constant help you provide through this forum.
Then, I am going to try to explain the difficulties I am finding when pre-processing DWI data. To start, I am going to describe the DWI protocol, since one of the main problems consist in the acquisition parameters.
Series UID: 1.3.12.2.1107.5.2.32.35461.2014110611304734506636119.0.0.0
Study date: 20141106
Study time: 104650.453000
Series date: 20141106
Series time: 113218.234000
Subject: GBRAIN003
Subject birth date: 19790110
Series description: ep2d_diff_mddw_20_p2
Manufacturer: SIEMENS
Model name: TrioTim
Software version: syngo MR B17
Study id: 1
Series number: 17
Repetition time (ms): 3300
Echo time[1] (ms): 90
Flip angle: 90
Number of averages: 1
Slice thickness (mm): 4
Slice spacing (mm): 5.199999931682
Image columns: 640
Image rows: 640
Phase encoding direction: COL
Voxel size x (mm): 1.79688
Voxel size y (mm): 1.79688
Number of volumes: 93
Number of slices: 25
Number of files: 93
Number of frames: 0
Slice duration (ms) : 0
Volume interval (s): 3.3125
Orientation: tra
Mosaic rows: 128
Mosaic columns: 128
Acquisition order: Interleaved Odd
Then, I would like to state that the protocol have three b=0 volumes, adquired all at the beginning of the protocol as the three first volumes, and with the same phase encoding direction (A >> P). Then, as I have understood from the fslwiki in the FDT, topup, eddy User Guides, and specifically in the eddy section advices before acquiring the data, it may be a problem not having acquired b=0 volumes with opposing PE directions, and not having acquired additional b=0 volumes intersperse through the main data set.
Additionally, I have tried applying topup:
fslroi GBRAIN0XX A2P_b0 0 3
printf "0 -1 0 0.095\n0 -1 0 0.095\n0 -1 0 0.095" > acqparams.txt
topup --imain=A2P_b0 --datain=acqparams.txt --config=b02b0.cnf --out=topup_output --fout=my_field --iout=unwarped_images --subsamp=1
And then eddy:
fslmaths unwarped_images -Tmean unwarped_images
bet unwarped_images unwarped_images_brain -m
indx=""
for ((i=1; i<=64; i+=1)); do indx="$indx 1"; done
echo $indx > index.txt
eddy --imain=GBRAIN003_ep2d_diff_mddw_20_p2 --mask=unwarped_images_brain_mask --acqp=acqparams.txt --index=index.txt --bvecs=GBRAIN003_ep2d_diff_mddw_20_p2_bvecs --bvals=GBRAIN003_ep2d_diff_mddw_20_p2_bvals --topup=topup_output --niter=8 --fwhm=10,8,4,2,0,0,0,0 --repol --out=eddy_corrected_data --mporder=6 --slspec=my_slspec.txt --s2v_niter=5 --s2v_lambda=1 --s2v_interp=trilinear
Although for the FSL version I am using in Ubuntu functions --repol and --mporder=6 are not included, so the final configuration was:
eddy --imain=GBRAIN003_ep2d_diff_mddw_20_p2 --mask=unwarped_images_brain_mask --acqp=acqparams.txt --index=index.txt --bvecs=GBRAIN003_ep2d_diff_mddw_20_p2_bvecs --bvals=GBRAIN003_ep2d_diff_mddw_20_p2_bvals --topup=topup_output --niter=8 --fwhm=10,8,4,2,0,0,0,0 --out=eddy_corrected_data
Finally, I have applied BET and dtifit:
bet eddy_corrected_data eddy_corrected_data_brain -m -R -f 0.2
dtifit --data=eddy_corrected_data --out=diffusion_tensor --mask=eddy_corrected_data_brain_mask --bvecs=GBRAIN003_ep2d_diff_mddw_20_p2_bvecs --bvals=GBRAIN003_ep2d_diff_mddw_20_p2_bvals --sse --wls --save_tensor
However, after revising each process' outputs, the images resulting from topup and eddy, visually, appear to be more distorted than the original images, and since --repol and --mporder=6 have not also been applied, I am not confident in the results from the tensor fitting.
Then, I do not know if there is any solution or adaptation from the original processing pipeline to process the data considering the specific characteristics from the acquisition protocol.
Then, I want to state that I have also tried the old eddy_correct function, but the output is an exact copy from the original image, so I can imagine that this is not the best option.
So, any advise or help to improve the processing pipeline and adapt it to the protocol characteristics will be extremely appreciated.
All the best,
Cristian MartÃnez Ruiz de Lara.
PhD Candidate.
Mind, Brain and Behavior Research Center.
University of Granada.
Spain.
|