Header

Poor DTI to T2ACPC alignment

Submitted by rlepping on
Category

I have a good AC-PC aligned T2star image for each subject that aligns well with the DTI for most subjects, but some DTI alignment with this T2star ACPC image is poor.  Are there alignment options within the DIFFPREP command to improve this?

 

Thanks!

Becca

Submitted by neon.taylor@gm… on Fri, 07/12/2019 - 13:32

Permalink

Hi-

 

Just to clarify: 

1) Does this mean that you are using a T2* (and not a T2w) volume as a reference volume in either DIFFPREP?

2) Are you aligning the T2* to your already-TORTOISE-processed data, and *that* is failing?

3) If you overlay the T2* volume on its unhappy partner DWI in some viewer (like AFNI!), do they overlay at all?

4) If you have AFNI, what is the output of the following command:

  3dinfo -o3 -extent -orient -prefix DWI_FILENAME  T2_FILENAME

?

 

--pt

 

1) Does this mean that you are using a T2* (and not a T2w) volume as a reference volume in either DIFFPREP?

Yes.  It was what was available.

 

2) Are you aligning the T2* to your already-TORTOISE-processed data, and *that* is failing?

I don't think so, but I'm not sure what the order of operations is within the DIFFPREP command.

 

3) If you overlay the T2* volume on its unhappy partner DWI in some viewer (like AFNI!), do they overlay at all?

They do overlay, just not as well aligned as with AFNI's align_epi_anat.py.

 

4) If you have AFNI, what is the output of the following command:

  3dinfo -o3 -extent -orient -prefix DWI_FILENAME  T2_FILENAME

3dinfo -o3 -extent -orient -prefix dti/sub_025_Control_dti.nii anat/sub_025_Control_T2_ss_acpc.nii
-154.880981    158.134521    -88.184845    -154.880981    142.775269    -139.521729    158.134521    -88.184845    59.815155    RPI           sub_025_Control_dti.nii
-90.164062    126.210899    -72.164062    -90.164062    91.164062    -91.210976    126.210899    -72.164062    109.164062    RPI    sub_025_Control_T2_ss_acpc.nii

 

Submitted by nayaka on Fri, 07/12/2019 - 17:21

Permalink

Hi Becca,

To further investigate the problem, if you are running DIFFPREP with EPI correction switched on, it would be helpful to run the command with the tag --keep_intermediate  1. This will write out all the outputs in the _proc folder and helps to understand the exact step where the processing is failing. My intuition is that the last step of DIFFPREP i.e rigid registration to structural maybe failing.

The main intermediate file of interest would be 'rpd_epi_b0.nii' and 'rpd_epi_structrual.nii'.  Please note that these files are not in the structural space yet but infact still in the b0 space. What these files will help understand is if the structural file you are providing is being used correctly by DIFFPREP. If the rpd files look aligned but the DMC.nii and DMC_structrual do not look aligned, then you can confirm that the last rigid step is failing. 

Amritha

Submitted by rlepping on Mon, 11/04/2019 - 12:18

In reply to by nayaka

Permalink

If the rpd files look aligned but the DMC.nii and DMC_structrual do not look aligned, then you can confirm that the last rigid step is failing. 


This is in fact what is happening.  I'm not seeing any options for the last rigid step in the example_registration_settings.dmc file or in the online help.  I was able to align the files with AFNI's align_epi_anat.py and feed the aligned files into DIFFPREP, but it would still be nice to know why it fails with DIFFPREP.

 

I am using a T2STAR rather than a T2 (it's what we had available).  Also, we do not have blip up and blip down data.  The output of the 3dinfo command is:

3dinfo -o3 -extent -orient -prefix dti/sub_025_Control_dti.nii anat/sub_025_Control_T2_ss_acpc.nii
-154.880981    158.134521    -88.184845    -154.880981    142.775269    -139.521729    158.134521    -88.184845    59.815155    RPI           sub_025_Control_dti.nii
-90.164062    126.210899    -72.164062    -90.164062    91.164062    -91.210976    126.210899    -72.164062    109.164062    RPI    sub_025_Control_T2_ss_acpc.nii

 

Thank you for your input!

Becca

Submitted by irfanogl on Mon, 11/04/2019 - 12:24

Permalink

Also just to add a little bit,

 

The version of TORTOISe that will be released this week has the registration to the structural image significantly improved.

That version might directly work.