ENH: Output thickness, curvature, and sulcal depth files

[ericfeczko]

This address issue #288. I'm running into some errors with the outputs due to some syntax issues that were introduced. Will be doing some testing with circle CI today to debug. Will also ensure pep8 compatibility.

[effigies]

If we restore all of the removed lines, we'll preserve the work done in #295. Then I think we can reduce this to a relatively small addition. We'll then need to go into outputs.py and add a couple nodes to write out the files correctly, mirroring what we do for .surf.gii:

smriprep/smriprep/workflows/outputs.py

Lines 592 to 601 in b29578e

	# Surfaces
	name_surfs = pe.MapNode(
	Path2BIDS(), iterfield="in_file", name="name_surfs", run_without_submitting=True
	)
	ds_surfs = pe.MapNode(
	DerivativesDataSink(base_directory=output_dir, extension=".surf.gii"),
	iterfield=["in_file", "hemi", "suffix"],
	name="ds_surfs",
	run_without_submitting=True,
	)

[effigies]

I think we're going to have trouble conflating .surf.gii files and .shape.gii files. I think we should probably not combine them and instead make a new morphometrics output. Something like:

Suggested change

	(fs2gii, allsurf_list, [('converted', 'in1')]),
	(fscurv2funcgii, allsurf_list, [('converted', 'in2')]),
	(allsurf_list, outputnode, [('out', 'surfaces')])
	(fscurv2funcgii, outputnode, [('converted', 'morphometrics')]),

[ericfeczko]

that makes a lot of sense -- I'll make sure to do the same moving forward when handling the cifti conversion

[effigies]

Suggested change

	allsurf_list = pe.Node(
	niu.Merge(7),
	name="allsurf_list",
	run_without_submitting=True,
	)

[effigies]

We don't need to iterate if in_file is always the same.

Suggested change

	whitesurf_list = pe.Node(
	niu.Merge(3, ravel_inputs=True),
	name="whitesurf_list",
	run_without_submitting=True,
	)
	fscurv2funcgii = pe.MapNode(
	fs.MRIsConvert(out_datatype="gii", to_scanner=True),
	iterfield=["in_file", "scalarcurv_file"], name="fscurv2funcgii",
	)
	fscurv2funcgii = pe.MapNode(
	fs.MRIsConvert(out_datatype="gii"),
	iterfield=["scalarcurv_file"], name="fscurv2funcgii",
	)

[ericfeczko]

won't the in_file vary by hemisphere? I'm a little confused by the interface, I'll admit :)

[effigies]

Oh right. That might be the best way to do this after all.

[effigies]

Suggested change

	(get_surfaces, whitesurf_list, [('smoothwm', 'in1'),
	('smoothwm', 'in2'),
	('smoothwm', 'in3')]),
	(whitesurf_list, fscurv2funcgii, [('out', 'in_file')]),
	(get_surfaces, fscurv2funcgii, [('smoothwm', 'in_file')]),

[effigies]

@ericfeczko The requirement that we find the right ?h.white (or ?h.smoothwm) for this is pushing logic that should be into the interface into the workflow. I think a cleaner solution is to make a new interface that just sub-classes MRIsConvert and infers the correct in_file if scalarcurv is passed. WDYT?

[ericfeczko]

that sounds perfect -- I'm in meeting hell right now, but I can try to give this a shot tonight or tomorrow.

[effigies]

Here's an (untested) attempt: 857e0bd

You may need to add my fork to try it out:

git remote add effigies https://github.com/effigies/smriprep.git
git fetch effigies
git cherry-pick 857e0bd

[effigies]

What are the expected outputs? I guess .dscalar.nii or .shape.gii?

[mgxd]

Closed by #305