Registration to T2W

50 views
Skip to first unread message

Shihan

unread,
Aug 4, 2021, 11:35:34 PMAug 4
to DSI Studio
Hi Frank,

Thank you for your software.

I'm currently using DSI Studio for fibre tractography in skeletal muscles. I have two questions about the automatic registration from DTI to T2W images.

1. I tried to upsample the DTI to the T2W space by using [Step T2][Edit][Resample to T1W/T2W]. The T2W with nii format was imported as a reference. However, the registration only fits some slices, as shown below. The registration results were poor at the wrist(fig1) and elbow(fig3). I also tried to run affine registration rather than rigid body, but the result was also not very well. Could you give me some advice on the processing steps or recommend some other tools to perform the registration?
registration_wrist.PNGregistration_middle.PNGregistration_elbow.PNG

2. Another question is that if I performed [Rotate to T1W/T2W] in [Step T2], the output fib file seems to have wrong fa values as shown in fig4, compared with the fa values without the preprocessing (fig5). And in fig4 condition, fibers can hardly be tracked.
fa_1.PNG
fa_2.PNG

Could you give me some advice on solving these issues? Thank you. Look forward to your reply.

Best regards,
Shihan

Frank Yeh

unread,
Aug 6, 2021, 3:52:01 PMAug 6
to dsi-s...@googlegroups.com
Hi Shihan,

    It is likely that the DWI has a large distortion that causes misalignment, and there is not much we can do other than correcting the distortion using FSL TOPUP.

    The second problem. You may update to the most recent version of DSI Studio and generate a new FIB file to see if it works.

Best regards,
Frank

--
You received this message because you are subscribed to the Google Groups "DSI Studio" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dsi-studio+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/52d9385d-afe7-400d-9584-830f300f1d17n%40googlegroups.com.

Shihan

unread,
Aug 8, 2021, 5:47:34 AMAug 8
to DSI Studio
Hi Frank,

Thanks for your reply. 

1-1 I found that the iteration time and registration type in the advanced settings influence the registration results. By using an affine registration and manually adjusting the scaling, shearing, and rotation parameters, I managed to obtain an acceptable result after running registration several times. The z scaling factor matters a lot. Do you think the manual adjustment is proper and necessary to get good registration?
1-2 The slice orientation of DWI is sagittal while the orientation of T2W is axial in our data. Could it be this orientation difference that leads to poor registration?

2-1 I downloaded the newest version of DSI Studio via 'dsi_studio_windows_64.zip'. However, the second problem still rises and the FA map seems to be almost "white" after opening the FIB file. I wondered if this results from the difference in the orientations between DWI and T2W sequences. Could I manually adjust the registration in [Step 3] after inserting the T2W images, rather than upsample the DWI data to T2W in [Step 2]? 
2-2 By the way, I tried to output tensor matrix in [Step 2] by adding txx, txy, txz, tyy, tyz, tzz in the 'Other output metrics'. However, the FIB file still does not contain the metrics of tensor matrix. Could you please give me some advice on getting the tensor matrix?

Really appreciate your help. Look forward to your reply.

Best regards,
Shihan

Frank Yeh

unread,
Aug 11, 2021, 10:39:13 PMAug 11
to dsi-s...@googlegroups.com
I am sorry for my late reply.


>
> 1-1 I found that the iteration time and registration type in the advanced settings influence the registration results. By using an affine registration and manually adjusting the scaling, shearing, and rotation parameters, I managed to obtain an acceptable result after running registration several times. The z scaling factor matters a lot. Do you think the manual adjustment is proper and necessary to get good registration?


Yes, because there is likely a distortion in the DWI.

>
> 1-2 The slice orientation of DWI is sagittal while the orientation of T2W is axial in our data. Could it be this orientation difference that leads to poor registration?


Likely, if the orientation matrix is not properly recorded in the
NIFTI header or DICOM header

>
>
> 2-1 I downloaded the newest version of DSI Studio via 'dsi_studio_windows_64.zip'. However, the second problem still rises and the FA map seems to be almost "white" after opening the FIB file. I wondered if this results from the difference in the orientations between DWI and T2W sequences. Could I manually adjust the registration in [Step 3] after inserting the T2W images, rather than upsample the DWI data to T2W in [Step 2]?


It is likely to be a bug in DSI Studio, but I could not reproduce the
same problem at my side.
Could you send me your SRC file and T2W to fix the program?

>
> 2-2 By the way, I tried to output tensor matrix in [Step 2] by adding txx, txy, txz, tyy, tyz, tzz in the 'Other output metrics'. However, the FIB file still does not contain the metrics of tensor matrix. Could you please give me some advice on getting the tensor matrix?
>

Please specify "tensor" to get these metrics.

Best regards,
Frank

>
> Really appreciate your help. Look forward to your reply.
>
> Best regards,
> Shihan
> On Saturday, 7 August 2021 at 03:52:01 UTC+8 Frank Yeh wrote:
>>
>> Hi Shihan,
>>
>> It is likely that the DWI has a large distortion that causes misalignment, and there is not much we can do other than correcting the distortion using FSL TOPUP.
>>
>> The second problem. You may update to the most recent version of DSI Studio and generate a new FIB file to see if it works.
>>
>> Best regards,
>> Frank
>>
>> On Wed, Aug 4, 2021 at 11:35 PM Shihan <mmass...@gmail.com> wrote:
>>>
>>> Hi Frank,
>>>
>>> Thank you for your software.
>>>
>>> I'm currently using DSI Studio for fibre tractography in skeletal muscles. I have two questions about the automatic registration from DTI to T2W images.
>>>
>>> 1. I tried to upsample the DTI to the T2W space by using [Step T2][Edit][Resample to T1W/T2W]. The T2W with nii format was imported as a reference. However, the registration only fits some slices, as shown below. The registration results were poor at the wrist(fig1) and elbow(fig3). I also tried to run affine registration rather than rigid body, but the result was also not very well. Could you give me some advice on the processing steps or recommend some other tools to perform the registration?
>>>
>>>
>>> 2. Another question is that if I performed [Rotate to T1W/T2W] in [Step T2], the output fib file seems to have wrong fa values as shown in fig4, compared with the fa values without the preprocessing (fig5). And in fig4 condition, fibers can hardly be tracked.
>>>
>>>
>>>
>>> Could you give me some advice on solving these issues? Thank you. Look forward to your reply.
>>>
>>> Best regards,
>>> Shihan
>>>
>>> --
>>> You received this message because you are subscribed to the Google Groups "DSI Studio" group.
>>> To unsubscribe from this group and stop receiving emails from it, send an email to dsi-studio+...@googlegroups.com.
>>> To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/52d9385d-afe7-400d-9584-830f300f1d17n%40googlegroups.com.
>
> --
> You received this message because you are subscribed to the Google Groups "DSI Studio" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to dsi-studio+...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/6b066a17-9003-4701-9951-f5796f0dde4dn%40googlegroups.com.

Shihan

unread,
Aug 22, 2021, 10:21:25 PMAug 22
to DSI Studio
Hi Frank,

Thank you for your reply!

Adding "tensor" in the advanced options works. I have uploaded a zip file named "data_white_fa_map.zip", in which the SRC file, a FIB file with upsampling to T2W, a FIB file without upsampling to T2, the T2W.nii, as well as the raw data of DTI and T2 are attached.

I wondered whether the "white" fa map might come from the great discrepancy of resolution between DTI (0.6154 pixels per mm with the matrix of 64*160) and T2W (3.7333 pixels per mm with the matrix of 448*448). Could this problem relate to the interpolation method that upsamples the DTI data to T2W?

By the way, I also want to check if my understanding of the z scaling in affine registration is correct. For example, if our DTI is sagittal with pixel size 1.625*1.625 mm, while the T2W is axial with slice spacing 3mm, should we adjust the z scaling parameter around 3/1.625 for a better affine registration?

Thanks again for your help!

Best regards,
Shihan

Frank Yeh

unread,
Aug 24, 2021, 5:15:43 PMAug 24
to dsi-s...@googlegroups.com
I will check and let you know.
Frank
> To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/87ba85ac-e1f9-4066-a203-37284cf027cbn%40googlegroups.com.

Frank Yeh

unread,
Sep 6, 2021, 1:20:32 PMSep 6
to dsi-s...@googlegroups.com
I'm sorry this issue takes so long because I haven't been able to find the cause yet.

A quick solution is to resample T2W to isotropic images using [Tools][O4: view image][Edit][Make isotropic]

This should give a correct FA calculation.

Best regards,
Frank

Frank Yeh

unread,
Sep 26, 2021, 11:00:59 PMSep 26
to dsi-s...@googlegroups.com
I fixed this issue. Please update DSI Studio (today's version) and see if it works for you now.
Thanks for reporting.
Frank
Reply all
Reply to author
Forward
0 new messages