Updated cortical thickness and curvature regression #300
Conversation
Co-authored-by: Tim Coalson <[email protected]>
Co-authored-by: Tim Coalson <[email protected]>
Co-authored-by: Tim Coalson <[email protected]>
Co-authored-by: Tim Coalson <[email protected]>
Co-authored-by: Tim Coalson <[email protected]>
Bunch of new scripts and some changes in postfreesurfer.
| @@ -70,6 +70,7 @@ opts_AddOptional '--inflatescale' 'InflateExtraScale' 'number' "surface inflatio | |||
| opts_AddOptional '--processing-mode' 'ProcessingMode' 'HCPStyleData|LegacyStyleData' "disable some HCP preprocessing requirements to allow processing of data that doesn't meet HCP acquisition guidelines - don't use this if you don't need to" 'HCPStyleData' | |||
| opts_AddOptional '--structural-qc' 'QCMode' 'yes|no|only' "whether to run structural QC, default 'yes'" 'yes' | |||
| opts_AddOptional '--use-ind-mean' 'UseIndMean' 'YES or NO' "whether to use the mean of the subject's myelin map as reference map's myelin map mean, defaults to 'YES'" 'YES' | |||
| opts_AddOptional '--metric-regression' 'MetricReg' 'OLD, NEW or BOTH' "whether to use the updated curvature-thickness regression, defaults to 'B=BOTH'" 'B' | |||
There was a problem hiding this comment.
Why do we want an option to not generate one of these? That would mean any later pipelines that use thickness and don't want curvature effects would have to be adapted to what the user chose for this setting.
I would think we want both to always be generated, without giving the user a choice to make their dataset more complicated for later handling.
There was a problem hiding this comment.
The old method is deprecated because it is substantially inferior. We are not using corrthickness much in pipelines (I think only the IDP pipeline) other than resampling it around.
There was a problem hiding this comment.
The old version is easy to compute and not a large file. I don't see an obvious downside to keeping backward compatibility. I assume we don't want to use the old filename for the new output. What exactly are you suggesting by "deprecated"?
There was a problem hiding this comment.
It's purpose is entirely superseded by the new version.
There was a problem hiding this comment.
Okay, and because of that...what do you want the pipelines to do about it?
There was a problem hiding this comment.
That is the current behavior, so we are providing a way to support that. We need to be able to make changes to already written pipelines. We can have transitional periods for deprecated features. corrThickness is obsolete with this new version. I don't really understand why this is so controversial. The pipelines need to be able to evolve with time.
There was a problem hiding this comment.
Huh? I'm not saying the pipelines shouldn't change, where did you get that? It is controversial because you aren't providing a reason for your strong position.
Why does the user need an option to not run the new method? It doesn't overwrite a previously-released filename with different contents, right? So the new files can't interfere with any old code (unless the user happened to already generate files with those new names), right? If so, then why do we want an "old behavior" mode, when the new behavior doesn't change any files that the previous version generated? Why would a user want that?
There was a problem hiding this comment.
Is the concern that someone won't like the added runtime or dependencies from the new version?
There was a problem hiding this comment.
It is quite a bit longer.
There was a problem hiding this comment.
People will generally only run PostFS after they have run FreeSurfer. How does it compare to the runtime of FreeSurfer?
Some simplification and cleaning; Including -metric-palette, -cifti-palette, and -metric-mask; Replacing "B" by "BOTH" in PostFreesurfer; Changing map names
Included resampling with both MSMSulc and MSMAll
global/scripts/CorrThick.sh
Outdated
| opts_AddMandatory '--subject-dir' 'SubjectDir' 'path' "folder containing all subjects" | ||
| opts_AddMandatory '--subject' 'Subject' 'subject ID' "subject-id" | ||
| opts_AddMandatory '--regname' 'RegName' 'my reg' "set the desired registration name(s) separated by @, 'string' 'RegName@RegName@[email protected].' use MSMSulc as default" | ||
| opts_AddOptional '--hemi' 'Hemi' 'hemisphere' "provide hemisphere for regression calculation, L=Left, R=Right, default 'L R' or B=Both" "B" |
There was a problem hiding this comment.
Why do we want the user to have the option not to run it on both hemispheres? Other scripts loop over a hardcoded L R.
Even if we keep it, the current description is disordered:
| opts_AddOptional '--hemi' 'Hemi' 'hemisphere' "provide hemisphere for regression calculation, L=Left, R=Right, default 'L R' or B=Both" "B" | |
| opts_AddOptional '--hemi' 'Hemi' 'hemisphere' "provide hemisphere for regression calculation, L=Left, R=Right, or B=Both, default 'B'" "B" |
New folder for all python scripts Sanity check for optional boolean argument
Changed --no-computation to --skip-computation
Bunch of new files and some changes in PostFreeSurfer.