warp fields are not created without this change and the workflow fails silently

[chrisgorgo]

No description provided.

[coveralls]

Coverage remained the same when pulling 33073f9 on chrisfilo:fix/apply_warp into cba8e41 on nipy:master.

[satra]

if this is required then the connections to warping stage are incorrect - it uses the field file. we need to check the workfow then.

[chrisgorgo]

The problem is that applywarp will not complain if it does not get a warp -
it will just resample the image. So even though there is a connection
nothing was passed along it because FNIRT did not produce the warp fields.
That's why this but went unnoticed.

On Mon, Apr 28, 2014 at 4:58 PM, Satrajit Ghosh notifications@github.comwrote:

if this is required then the connections to warping stage are incorrect -
it uses the field file. we need to check the workfow then.

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/838#issuecomment-41568362
.

[satra]

i mean the field_file is different from the fieldcoeff_file and it was passing along the field_file which may be incorrect!

[satra]

so we should check the workflow to see what the outputs look like.

[mwaskom]

Ah, can confirm that I've had to do this in my warp workflow too: https://github.com/mwaskom/lyman/blob/master/lyman/workflows/anatwarp.py#L97

[thelxinoe]

I'm not sure whether another bug or not but I just wanted to point out that the fsl registration in the workflows is wrong as well:

nipype/nipype/workflows/fmri/fsl/preprocess.py

Line 1171 in b1b7825

anat2target_nonlinear = pe.Node(fsl.FNIRT(), name='anat2target_nonlinear')

They all use field and none of the warping is instigated.

[satra]

@thelxinoe - i believe it's a bug - since i originally wrote that bit, i will look into it tomorrow and fix it.

[chrisgorgo]

To avoid such mistakes in the future maybe we should make "field" input of ApplyWarp mandatory and use ApplyXfm instead of ApplyWarp for linear only transformations?

[thelxinoe]

I fixed this, I'm using the workflow, just by turning on fieldcoeff_file = True and replacing field_file with fieldcoeff_file.

I also noticed that there is a mistake at the beginning of the same workflow where the non-bet structural image is used for mean registration. This doesn't really work.

nipype/nipype/workflows/fmri/fsl/preprocess.py

Line 1143 in b1b7825

register.connect(inputnode, 'anatomical_image', mean2anat, 'reference')

Should I submit them?

[chrisgorgo]

Please do. Thank you for your contributions. We have found the same
BET/FLIRT issue in BIPs - will fix it too.

On Thu, May 1, 2014 at 12:29 PM, thelxinoe notifications@github.com wrote:

I fixed this, I'm using the workflow, just by turning on fieldcoeff_file =
True and replacing field_file with fieldcoeff_file.

I also noticed that there is a mistake at the beginning of the same
workflow where the non-bet structural image is used for mean registration.
This doesn't really work.

nipype/nipype/workflows/fmri/fsl/preprocess.py

Line 1143 in b1b7825

register.connect(inputnode, 'anatomical_image', mean2anat, 'reference')

I'm just checking that they work but should I submit them?

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/838#issuecomment-41898493
.

[thelxinoe]

Okay, I noticed a couple of other faults I'll submit them in a bit. Thanks.

[thelxinoe]

Okay, I found a number of bugs with respects to FSL and the application of certain affine matrices. It seems they can only be applied with flirt and not applywarp or applyxfm4d.

I'll rewrite tomorrow to reflect this.

[chrisgorgo]

Thanks!

On Thu, May 1, 2014 at 7:05 PM, thelxinoe notifications@github.com wrote:

Okay, I found a number of bugs with respects to FSL and the application of
certain affine matrices. It seems they can only be applied with flirt and
not applywarp or applyxfm4d.

I'll rewrite tomorrow to reflect this.

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/838#issuecomment-41930713
.

[satra]

any updates on this?

[thelxinoe]

I think the fix has been accepted?
On 15 Jul 2014 01:44, "Satrajit Ghosh" notifications@github.com wrote:

any updates on this?

—
Reply to this email directly or view it on GitHub
#838 (comment).

[satra]

closing this one. this fax has been accepted and a newer set of patches is now available #877