ENH: *fieldmap-less* integration from *SDCFlows* 2.0 #153
Conversation
|
Hello @oesteban, Thank you for updating! Cheers! There are no style issues detected in this Pull Request. 🍻 To test for issues locally, Comment last updated at 2021-03-11 13:53:27 UTC |
oesteban
force-pushed
the
enh/fieldmap-less
branch
from
61e3c49
to
b6bda36
Compare
oesteban
force-pushed
the
enh/fieldmap-less
branch
2 times, most recently
from
cbdbc99
to
49d511b
Compare
oesteban
force-pushed
the
enh/fieldmap-less
branch
2 times, most recently
from
8a8e996
to
c50afb9
Compare
dmriprep/workflows/dwi/base.py
Outdated
| @@ -297,7 +303,10 @@ def _bold_reg_suffix(fallback): | |||
| ) | |||
|
|
|||
| sdc_report = pe.Node( | |||
| SimpleBeforeAfter(before_label="Distorted", after_label="Corrected",), | |||
| SimpleBeforeAfter( | |||
| before_label="Distorted", | |||
There was a problem hiding this comment.
The corrected image looks more distorted. Have the labels been swapped?
There was a problem hiding this comment.
What dataset are you looking at? both?
There was a problem hiding this comment.
Okay, I had a look and I believe this is a problem of the sloppy mode. The final B-Spline regularizer has very low resolution, and hence the distortions are converted into a very smooth field. It does a decent job to reshape the CC in the middle of the brain, but very bad job at the brain edges where the frequency of the distortion is high.
If you look at ds0001771, the reports are the same as in master, right?
If you look at ds000206, the fieldmap-less correction is pretty bad - here I think the sloppy mode is biting at all fronts. It is also that distortion is just too heavy.
Let's go ahead and try to refine this in the future. I think it is important not to be held up by optimization. There's a lot of integration ahead and all these moving pieces will need to find their right place. If we are too strict with that fitting right now we will not have a chance to look at the whole thing.
Currently, this is blocked by the absence of a general solution to the reference generation issue (see nipreps/niworkflows#601). Once adequate DWI references are connected to the estimation workflow, this feature should work out. We can test the new feature with ds000206. Resolves: #111.
oesteban
force-pushed
the
enh/fieldmap-less
branch
from
b251f17
to
71e9568
Compare
oesteban
force-pushed
the
enh/fieldmap-less
branch
from
71e9568
to
cfd1429
Compare
Currently, this is blocked by the absence of a general solution to the
reference generation issue (see nipreps/niworkflows#601).
Once adequate DWI references are connected to the estimation workflow,
this feature should work out.
We can test the new feature with ds000206.
Resolves: #111.