SDC-pepolar: topup implementation #76

markushs · 2019-12-13T11:00:07Z

Here's a solution that works for us (ref #37).

topup.py is a modification of pepolar.py and reuses many of the functions/workflows found in pepolar.py.

Note, I had to decrease the bet threshold in gre.py as the epis (which are used as magnitude inputs to fmap_wf) already got skullstripped in topup/init_prepare_epi_wf. This is probably not the best solution, as the relevant BET node is also used by other workflows with non-skullstripped mag-images. However I see that there's an ongoing discussion about changing the approach to something more robust than BET (#23).

pep8speaks · 2019-12-13T11:00:12Z

Hello @markushs, Thank you for updating!

Cheers! There are no style issues detected in this Pull Request. 🍻 To test for issues locally, pip install flake8 and then run flake8 sdcflows.

Comment last updated at 2019-12-18 08:45:03 UTC

mattcieslak · 2019-12-13T13:13:37Z

Thank you for starting this PR! I was wondering if you have run into any left-right flips between the fieldmaps and the epi data. Topup in some versions of FSL assumes that the image orientation is LAS+. Since fmriprep uses RAS+ this might be difficult to spot.

markushs · 2019-12-13T14:40:35Z

@mattcieslak I haven't seen any left-right flips, no. And visually inspecting epi data pre-post topup (i.e. distorted epi input and undistorted epi output) + inspecting headers suggest no flips have occured. Any bullet-proof way to test if something like this has occured?

oesteban · 2019-12-13T22:02:48Z

I was wondering if you have run into any left-right flips between the fieldmaps and the epi data.

Are you referring to inconsistency between them? i.e., one of them has an x-form header where there is an X-flip and the data are therefore interpreted in reversed way.

Topup in some versions of FSL assumes that the image orientation is LAS+.

Assumes LAS+ or assumes the same orientation information for both images? The former would be really dangerous, the latter would be just unfortunate but understandable from a historical point of view.

markushs · 2019-12-18T08:51:30Z

I did some additional testing, and with our data it looks like topup works better when epi fieldmaps are not skullstripped. The latest commit to my branch reflects this. As the "magnitude" input to fmap.py now contains skull, the frac-threshold in gre.py could be reverted to its original value (0.6).

oesteban · 2020-02-20T22:04:15Z

Hi @markushs, is there anything we can do to help you finalize this?

markushs · 2020-02-26T18:38:18Z

Hi @oesteban, have you had a chance to look at the draft? If the approach seems ok I can try to fix the remaining problems (some tests that need to be updated and an issue with the _desc_fieldmap_bold.svg, If I recall correctly). The skullstripping could also be refined, but that seems to be work in progress (ref #23)

oesteban

Getting there! Very looking forward to seeing this finalized :)

oesteban · 2020-02-26T19:17:37Z

sdcflows/workflows/base.py

-Based on the estimated susceptibility distortion, a corrected
-EPI (echo-planar imaging) reference was calculated for a more
-accurate co-registration with the anatomical reference.
+Based on the estimated susceptibility distortion, a corrected EPI (echo-planar imaging) reference


Please rebase or merge with current upstream so differences in this file are real.

As you can see I've opened a new PR (#106 based on the current master). Will respond to your comments here, but the perhaps better to continue the discussion in the new PR? Sorry for the mess, far from fluent in git...

oesteban · 2020-02-26T19:17:59Z

sdcflows/workflows/base.py

    if 'epi' in fmaps:
-        from .pepolar import init_pepolar_unwarp_wf, check_pes
+        from .topup import init_topup_wf, check_pes


I don't love the topup name of the module. I would prefer this workflow to be within the pepolar module.

Fixed in #106

oesteban · 2020-02-26T19:24:21Z

sdcflows/workflows/topup.py

+echo-planar imaging (EPI) references with opposing phase-encoding
+directions, with `Topup`).


Suggested change

echo-planar imaging (EPI) references with opposing phase-encoding

directions, with `Topup`).

echo-planar imaging (EPI) references of opposed phase-encoding

directions with `topup` @topup (FSL {fsl_ver}).

We will need to add a topup entry to the bib file, referencing

J.L.R. Andersson, S. Skare, J. Ashburner. How to correct susceptibility distortions in spin-echo echo-planar images: application to diffusion tensor imaging. NeuroImage, 20(2):870-888, 2003.

FSL version gets returned in #106, but not sure where I find the bib-file?

oesteban · 2020-02-26T19:25:21Z

sdcflows/workflows/topup.py

+    return workflow
+
+
+def init_prepare_epi_wf(omp_nthreads, matched_pe=False,


Can you unpack for me how much of this workflow overlaps the existing analog?

In #106 all edits were done in pepolar.py, so should be easier to evaluate the overlap

oesteban · 2020-02-26T19:26:01Z

sdcflows/workflows/topup.py

+    return inlist
+
+
+def check_pes(epi_fmaps, pe_dir):


oesteban · 2020-02-26T19:26:33Z

sdcflows/workflows/topup.py

+    return matched_pe, matched_pe_dir, opposed_pe_dir
+
+
+def _split_epi_lists(in_files, pe_dir, max_trs=50):


oesteban · 2020-02-26T19:29:59Z

sdcflows/workflows/base.py

+        sdc_unwarp_wf = init_sdc_unwarp_wf(
+            omp_nthreads=omp_nthreads,
+            debug=debug)


This is not a responsibility of this workflow. Instead, the warpfield should be converted to an ITK-type of transform and generate a new (unwarped) reference, as in:

sdcflows/sdcflows/workflows/pepolar.py

Lines 128 to 140 in bf9cd97

to_ants = pe.Node(niu.Function(function=_fix_hdr), name='to_ants',

mem_gb=0.01)

cphdr_warp = pe.Node(CopyHeader(), name='cphdr_warp', mem_gb=0.01)

unwarp_reference = pe.Node(ANTSApplyTransformsRPT(dimension=3,

generate_report=False,

float=True,

interpolation='LanczosWindowedSinc'),

name='unwarp_reference')

enhance_and_skullstrip_bold_wf = init_enhance_and_skullstrip_bold_wf(

omp_nthreads=omp_nthreads)

But that's going to be tricky while https://github.com/poldracklab/nitransforms does not support FSL's bspline parametric fields.

Not sure what you mean by this not being a responsibility of this workflow. I tried to summarize the approach in the comment to (#106), but essentially the topup-estimated fieldmap is fed into the workflows of the fmap-module.

markushs · 2020-06-18T07:29:23Z

Hi,
Sorry for the delay here (has been crazy times...)
I have some time now to finalize this, if it's still of interest @oesteban?
As I have to re-familiarize myself with the code (including my own) I was wondering if it would be cleaner to open a new PR based on the current master?

oesteban · 2020-06-22T15:33:16Z

I'll have a closer look later today or tomorrow. Since this PR seemed a bit abandoned, I made some progress on this front here: https://github.com/nipreps/dmriprep/pull/97/files. Hopefully, we can find the common points of that PR to dmriprep and this one to come up with a solution that works well regardless of the input dataset.

oesteban · 2020-11-18T13:29:03Z

Let's move the conversation to #106.

markushs added 3 commits December 12, 2019 17:21

topup implementation of pepolar workflow

c86d4d1

connect topup wf with FIELDMAP path

277533c

decrease bet frac-threshold

94e7bff

fixes flake8 comments

cb2afbe

markushs changed the title ~~Topup implementation of PEPOLAR flow~~ SDC-pepolar: topup implementation Dec 13, 2019

Adapted topup_mod: topup on nonskullstripped input

b3ab347

mgxd force-pushed the master branch from 2868c61 to bf9cd97 Compare February 15, 2020 18:17

oesteban reviewed Feb 26, 2020

View reviewed changes

markushs mentioned this pull request Jun 22, 2020

SDC-pepolar: topup implementation v2 #106

Closed

oesteban closed this Nov 18, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

SDC-pepolar: topup implementation #76

SDC-pepolar: topup implementation #76

markushs commented Dec 13, 2019

pep8speaks commented Dec 13, 2019 •

edited

Loading

mattcieslak commented Dec 13, 2019

markushs commented Dec 13, 2019

oesteban commented Dec 13, 2019

markushs commented Dec 18, 2019

oesteban commented Feb 20, 2020

markushs commented Feb 26, 2020

oesteban left a comment

oesteban Feb 26, 2020

markushs Jun 22, 2020

oesteban Feb 26, 2020

markushs Jun 22, 2020

oesteban Feb 26, 2020

markushs Jun 22, 2020

oesteban Feb 26, 2020

markushs Jun 22, 2020

oesteban Feb 26, 2020

markushs Jun 22, 2020

oesteban Feb 26, 2020

markushs Jun 22, 2020

oesteban Feb 26, 2020

markushs Jun 22, 2020

markushs commented Jun 18, 2020

oesteban commented Jun 22, 2020

oesteban commented Nov 18, 2020

		echo-planar imaging (EPI) references with opposing phase-encoding
		directions, with `Topup`).

		return workflow


		def init_prepare_epi_wf(omp_nthreads, matched_pe=False,

		return matched_pe, matched_pe_dir, opposed_pe_dir


		def _split_epi_lists(in_files, pe_dir, max_trs=50):

	to_ants = pe.Node(niu.Function(function=_fix_hdr), name='to_ants',
	mem_gb=0.01)

	cphdr_warp = pe.Node(CopyHeader(), name='cphdr_warp', mem_gb=0.01)

	unwarp_reference = pe.Node(ANTSApplyTransformsRPT(dimension=3,
	generate_report=False,
	float=True,
	interpolation='LanczosWindowedSinc'),
	name='unwarp_reference')

	enhance_and_skullstrip_bold_wf = init_enhance_and_skullstrip_bold_wf(
	omp_nthreads=omp_nthreads)

SDC-pepolar: topup implementation #76

SDC-pepolar: topup implementation #76

Conversation

markushs commented Dec 13, 2019

pep8speaks commented Dec 13, 2019 • edited Loading

Comment last updated at 2019-12-18 08:45:03 UTC

mattcieslak commented Dec 13, 2019

markushs commented Dec 13, 2019

oesteban commented Dec 13, 2019

markushs commented Dec 18, 2019

oesteban commented Feb 20, 2020

markushs commented Feb 26, 2020

oesteban left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

markushs commented Jun 18, 2020

oesteban commented Jun 22, 2020

oesteban commented Nov 18, 2020

pep8speaks commented Dec 13, 2019 •

edited

Loading