Re: [HCP-Users] gradient nonlinearity correction question

[Kim, Joo-won]

Hi Kristian,

Which one looks more reasonable, online or offline correction for subject data?

On my data, dicom header (0019, 1014) contains table position.

Best,
Joo-won

On 4/11/18, 9:31 AM, "Kristian Loewe" <k...@kristianloewe.com> wrote:

Hi Joo-won and Keith,

I don't think that the table has been moved. Is there any information  
somewhere in the dicom header to double-check this?

I am using the coeff_AS82.grad file for the Prisma data (the second  
command). The first command was what I used for the Verio data. Also,  
I double-checked that I am using the uncorrected volume as input.

I am not sure if I am supposed/allowed to send screenshots of the  
actual data in the subject's native space to the list. I'm going to  
check that. Meanwhile, I asked our local MR team to acquire an  
additional data set using a phantom. The difference between the  
scanner-corrected image and the offline-corrected image is not as  
striking in this case but it's visible. Based on the phantom data, I  
am inclined to say that both corrections work reasonably well, even  
though they are not exactly the same. I also tried the  
offline-correction on some phantom EPI data and it seems to work well  
as it (together with EPI distortion correction) restores the original  
shape of the phantom pretty nicely.

The differences between the correction variants applied to the  
subject's data are actually rather small inside the brain but become  
larger towards the neck, which is to be expected as the distance to  
the magnetic isocenter becomes greater in that direction.  
Nevertheless, I orginally thought that the differences between  
scanner- and offline-corrected images would be smaller than that.

Find attached some plots of the T1 phantom data:

T1_ND*.png - uncorrected images
T1*.png- scanner-corrected images
T1_ND_gdc*.png - offline-corrected images

Cheers,

Kristian


PS:
I am sending this email for the second time (apparently the attachment  
was too large). I am not sure if the first email was successfully  
cancelled. I apologize if you are receiving this twice now.

    
    Quoting "Kim, Joo-won" <joo-won@mssm.edu>:

> Hi Kristian,
>
> Have you moved table? If you moved the table, you should manually  
> subtract it from qform, sform, and/or affine matrix in the nifty  
> header.
>
> Best,
> Joo-won
>
> ---
> Joo-won Kim, Ph.D.
> Postdoctoral Fellow
> Translational and Molecular Imaging Institute
> Icahn School of Medicine at Mount Sinai
>
>
> From: <hcp-users-boun...@humanconnectome.org> on behalf of Keith  
> Jamison <kjami...@umn.edu>
> Date: Monday, April 9, 2018 at 10:36 AM
> To: Kristian Loewe <k...@kristianloewe.com>
> Cc: HCP Users <hcp-users@humanconnectome.org>
> Subject: Re: [HCP-Users] gradient nonlinearity correction question
>
> First make sure you're using the right coefficient file, copied  
> directly from the scanner.  The Prisma should have a file called  
> coeff_AS82.grad, so the one you used in your *second* command above  
> should be correct.
> Second is to be absolutely sure your input is the uncorrected volume 
(*_ND).
> If you include some matched screenshots of the uncorrected,  
> offline-corrected, and scanner-corrected volumes, we can maybe help  
> evaluate the difference.
>
>
> On Mon, Apr 9, 2018 at 4:09 AM, Kristian Loewe  
> <k...@kristianloewe.com<mailto:k...@kristianloewe.com>> wrote:
> Thanks Keith,
>
> Cropping is turned off by default in the version of dcm2niix that  
> I'm using but I re-ran the conversion with "-x n" anyway. I also  
> used fslreorient2std as per your suggestion. Unfortunately, the  
> result is still the same. Do you have any other ideas?
>
>
> Cheers,
>
> Kristian
>
>
> Quoting Keith Jamison <kjami...@umn.edu<mailto:kjami...@umn.edu>>:
> Some problems can arrise if the NIFTI files are unexpectedly manipulated
> prior gradient_unwarp. Two things to check:
>
> 1. dcm2nii and dcm2niix has options to perform additional processing like
> reorienting or cropping, some of which may be enabled by default. Make 
sure
> those are all DISABLED. (for dcm2niix add "-x n" and for dcm2nii you can
> add "-x N -r N"
> 2. We also usually use "fslreorient2std  _new and then
> 
https://urldefense.proofpoint.com/v2/url?u=http

Re: [HCP-Users] gradient nonlinearity correction question

[Kim, Joo-won]

Hi Kristian,

Have you moved table? If you moved the table, you should manually subtract it 
from qform, sform, and/or affine matrix in the nifty header.

Best,
Joo-won

---
Joo-won Kim, Ph.D.
Postdoctoral Fellow
Translational and Molecular Imaging Institute
Icahn School of Medicine at Mount Sinai


From:  on behalf of Keith Jamison 

Date: Monday, April 9, 2018 at 10:36 AM
To: Kristian Loewe 
Cc: HCP Users 
Subject: Re: [HCP-Users] gradient nonlinearity correction question

First make sure you're using the right coefficient file, copied directly from 
the scanner.  The Prisma should have a file called coeff_AS82.grad, so the one 
you used in your *second* command above should be correct.
Second is to be absolutely sure your input is the uncorrected volume (*_ND).
If you include some matched screenshots of the uncorrected, offline-corrected, 
and scanner-corrected volumes, we can maybe help evaluate the difference.


On Mon, Apr 9, 2018 at 4:09 AM, Kristian Loewe 
> wrote:
Thanks Keith,

Cropping is turned off by default in the version of dcm2niix that I'm using but 
I re-ran the conversion with "-x n" anyway. I also used fslreorient2std as per 
your suggestion. Unfortunately, the result is still the same. Do you have any 
other ideas?


Cheers,

Kristian


Quoting Keith Jamison >:
Some problems can arrise if the NIFTI files are unexpectedly manipulated
prior gradient_unwarp. Two things to check:

1. dcm2nii and dcm2niix has options to perform additional processing like
reorienting or cropping, some of which may be enabled by default. Make sure
those are all DISABLED. (for dcm2niix add "-x n" and for dcm2nii you can
add "-x N -r N"
2. We also usually use "fslreorient2std  _new and then
gradient_unwarp.py
 on _new

-Keith


On Fri, Apr 6, 2018 at 12:05 PM, Kristian Loewe 
>
wrote:
Hi,

I would like to use gradunwarp for offline gradient nonlinearity
correction of some data acquired on our local Siemens scanners. I used
dcm2niix to convert the dicom data to nifti format. After applying
gradunwarp to a T1 image in nifti format (the one that originally has
the _ND suffix), I proceeded to compare the result with the
Siemens-corrected T1 image. I expected that they would look very
similar but in fact they look quite different. I am wondering if this
is to be expected to some degree because of differences in the
correction algorithms or what else might be the reason for this. Could
it be the case, for example, that the wrong center is being used for
some reason?

I have tried this for a T1 image acquired on a Verio and another one
from a Prisma using

   
gradient_unwarp.py
 T1.nii.gz T1_gdc.nii.gz siemens -g coeff_AS097.grad
-n

and

   
gradient_unwarp.py
 T1.nii.gz T1_gdc.nii.gz siemens -g coeff_AS82.grad -n

respectively.

I would really appreciate any help or advice you can provide.

Cheers,

Kristian


Quoting Keith Jamison >:

> FYI, when available, you can enable it on the scanner in the
> "Resolution->Filter" tab with the "Distortion Correction" checkbox.  It's
> often used for structural scans like MPRAGE, where you will see two DICOM
> folders in the output:  and _ND.  ND means "No
> Distortion [Correction]".. .A very confusing choice of acronym.  You can
> then compare the online corrected (not _ND) and offline using gradunwarp.
>
> -Keith
>
>
> On Wed, Oct 19, 2016 at 4:44 PM, Glasser, Matthew 
> >
> wrote:
>
>> Some do for some sequences, but because it is not uniformly applied and
>> because they are likely not to use optimal interpolation algorithms, we
>> prefer to do offline correction.
>>
>> Peace,
>>
>> Matt.
>>
>> From: 
>> >
>>  on behalf of Antonin
Skoch <
>> a...@ikem.cz>
>> Date: Wednesday, October 19, 2016 at 4:27 PM
>> To: "hcp-users@humanconnectome.org" 
>> 

[HCP-Users] Transformation matrices of HCP 500 Subject Release

[Kim, Joo-won]

Dear HCP,

Thank you for the HCP data. I have a request.

Could you release the transformation matrices from unprocessed images to ACPC 
aligned images of the HCP 500 Subjects Release? I would like to analyze both 
averaged and individual scans, but when I ran the PreFreeSurfer, the 
ACPC-aligned images were shifted a little bit compared to those in the released 
data. Specifically, could you provide the following files corresponding to the 
subjects in the attachment?

T1w/AverageT1wImages/ToHalfTrans000?.mat or 
T1w/AverageT1wImages/ImToHalf000?.nii.gz
T1w/xfms/acpc.mat
T2w/AverageT2wImages/ToHalfTrans000?.mat or 
T2w/AverageT2wImages/ImToHalf000?.nii.gz
T2w/xfms/acpc.mat
T2w/T2wToT1wReg/T2w2T1w.mat

Best,
Joo-won


-- 
Joo-won Kim

Postdoctoral Fellow
Translational and Molecular Imaging Institute
Department of Radiology
Icahn School of Medicine at Mount Sinai

___
HCP-Users mailing list
HCP-Users@humanconnectome.org
http://lists.humanconnectome.org/mailman/listinfo/hcp-users
100307
100408
101107
101309
101915
103111
103414
103818
105014
105115
106016
108828
110411
111312
111716
113619
113922
114419
115320
116524
117122
118528
118730
118932
120111
122620
131217
153025
154734
188347
211720
212318
214423
221319
239944
245333
280739
298051
366446
397760
414229
499566
654754
672756
751348
756055
792564
856766
857263
899885