DTI preprocessing Qiang Wang Movement Disorders and Neuromodulation

DTI pre-processing Qiang Wang Movement Disorders and Neuromodulation Unit 26. 02. 2019

A common preprocessing pipeline for DTI preprocessing 1. DICOMs to NIf. TIs 2. Topup: correcting Susceptibility-induced Distortions 3. Eddy: correcting Eddy Currents distortions and head movement

Topup: Susceptibility-induced Distortions, caused by difference of gradient field; most interfaces between air, bone, and soft tissues, Green: real brain

Eddy Currents: generated by conductive material coils, shields, tubes, wires or devices within or on the patient

tools Dcm 2 niix: convert DICOMs to NIFT files, advantages: 1. Ni. FTs and bvec are compatible with FSL 2. b-vec will be oriented w FSL: pre-processing

Diffusion data: • dti. nii(4 D nifts): dti_AP. nii , Anterior Posterior phase-encoding direction acquisitions. dti_PA. nii , Posterior Anterior PE acquisitions Or RL and LR acquisitions. • bval: short for b value, is a factor that reflects the strength and timing of the gradients used to generate diffusion-weighted images • bvec: direction. short for b vectors, contains a 3 x 1 vector for each gradient, x, y, z, indicating the gradient PS: if no complete PA acquisitions, additional PA phase-encoded b 0 image at least

b=0 b=1000 Brain DWI images using 3 different b-values (0, 1000, and 3000) b=3000

dti. nii bval=1000, n=69 Fslinfo dti_AP. nii bval=0, n=1 dimension volumes spacing

bval bvec

Topup 1. extract a volume without diffusion weighting from dti. nii, b 0 (AP PE direction) and b 0 (PA PE direction) fslroi dti_AP nodif_AP 0 1 fslroi dit_PA nodif_PA 0 1 2 an acqparams. txt, containing the following rows: -1 means AP direction, 0 -1 0 0. 075184 1 means PA direction, 0. 075184 is readout time 0 1 0 0. 075184 3. merge the AP(nodif_AP. nii) and PA(nodif_PA. nii) images into a single image fslmerge -t AP_PA_b 0 nodif_PA 4. topup: topup --imain=AP_PA_b 0 --datain=acqparams. txt --config=b 02 b 0. cnf --out=topup_AP_PA_b 0 --iout=topup_AP_PA_b 0_fout --fout=topup_AP_PA_b 0_iout (movement between nodif_AP and nodif_PA) (Susceptibility information off-resonance field) (off-resonance field looks like) ( unwarped and corrected image)

topup before after

Topup 5. applytopup: apply corrected distortion by topup for applytopup --imain=dti_AP -inindex=1 -datain=acqparams. txt -topup=topup_AP_PA_b 0 -out=hifi_dti --method=jac acqparams. txt (Corrected distortions dti_AP acquisition) 0 -1 0 0. 075184 (AP direction) 0 1 0 (PA direction) 0. 075184 For example: imain=dit_PA--, inindex=2 -- remaining volumes (AP or/and PA direction) ……

applytopup before dti_AP. nii(a volume, b=1000) after hifi_AP. nii(a volume, b=1000)

Eddy for eddy current and head movement 1. average corrected b 0 volumes(as a template) fslmaths topup_AP_PA_b 0_iout -Tmean hifi_nodif (Averaged AP and PA b 0 acquisitons) 2. creat a mask(based on hifi_nodif) to separate brain from nonbrain bet hifi_nodif_brain -m -f 0. 2 3. eddy --imain=dti_AP -mask=hifi_nodif_brain_mask -index=index. txt --acqp=acqparams. txt -bvecs=bvec --bvals=bval topup=topup_AP_PA_b 0 out=eddy_unwarped_images -- flm=quadratic ---fwhm=0 -data_is_shelled index. txt, contains a single row of ’ 1’, each 1 for each volume in dti_AP

DTI preprocessing before (dti_AP, b=0) (eddy_unwarped_images, b=0) after

DTI preprocessing before (dti_AP, b=1000) (eddy_unwarped_images, b=1000) after

Thanks for your attention !