search

pnlbwh / multi-shell-dMRIharmonization

Harmonization of multi-shell diffusion MRI, an enhancement of https://github.com/pnlbwh/dMRIharmonization
Other
8 stars 4 forks source link

Lower bshell measures being directly warped to MNI instead of template-->MNI #8

Closed tashrifbillah closed 4 years ago

tashrifbillah commented 4 years ago

It has been observed that lower bvalue-shell diffusion and rish measures are being directly warped to MNI space instead of being warped to template space first and then to MNI space.

Reason:

  1. In the first pass, for bmax-shell, only prefix*FA*1Warp.nii.gz transforms are present that properly warps subject space to template space.

  2. Later during debugging, template space is being warped to MNI space using prefix*FA*ToMNI1Warp.nii.gz

  3. However, in the second pass, with pattern prefix*FA*1Warp.nii.gz, there are two transforms present: prefix*FA*1Warp.nii.gz and prefix*FA*ToMNI1Warp.nii.gz. Thus, for lower bvalue-shells, scale maps are indeed created from MNI space data, not the template space data. This error propagated down the pipeline and resulted in bad result, again for lower bvalue-shells only.

Refer to codes: https://github.com/pnlbwh/multi-shell-dMRIharmonization/blob/c7ce6a4412bd7e91332949fcf170788d13796809/lib/buildTemplate.py#L50

and

https://github.com/pnlbwh/multi-shell-dMRIharmonization/blob/c7ce6a4412bd7e91332949fcf170788d13796809/lib/debug_fa.py#L64

tashrifbillah commented 4 years ago

Issue illustration by @yrathi

image

tashrifbillah commented 4 years ago

Only one word has been changed in the commit:

https://github.com/pnlbwh/multi-shell-dMRIharmonization/commit/f5346e34d145778415d6db24cc938fc8dac1da71

tashrifbillah commented 4 years ago

Please use temporary branch: https://github.com/pnlbwh/multi-shell-dMRIharmonization/tree/fix-warp

Again, it's already set up in our platforms.