I have some questions of some warning messages I got during creation of a population template. I want to create a template out of 50 subjects using TORTOISE 3.2 in Ubuntu 20.04. The command I used is follows: dtireg_create_template -s=sus.txt -n=1 -c=1 -r=1.5x1.75x2.5. Though the procedure ran through and displayed the completion message, I also got several warnings as shown below:
If create a template out of 4 subjects , these warnings will disappear completely.
Q2) I am glad it is solved.
Q3) This should be fine normally unless either DIPY or the conversion tool outputs the 4th dimension as 0. So what I suggest is to run the TORTOISE executable Set4DImageResolution.
If the original tensor is aaa_DT.nii with resolution 1.5x1.75x2.5x0, you run it by:
Set4DImageResolution aaa_DT.nii aaa_DT.nii 1.5 1.75 2.5 1 (last parameter is 1 instead of 0)
Hopefully, after this they should be compatible.
Q1) Unfortunately, DRTAMAS is not designed for anisotropic resolution. So it will always output at isotropic resolution so I storngly suggest to use "-r 1.5"
However, after the template is created, you can resample it with the command ResampleTensor.
You can run it by: ResampleImage average_template_diffeo_6.nii 1.5 1.75 2.5
and your new template will have the desired resolution.
Q4) Apologies but I did not fully understand your last question, What do you mean by "split the ROI area?
Hope these help.
The number of subjects should not matter. We have created templates out of 1000s of subjects.
I see a couple of issues here. (Some related, some are not)
1) DRTAMAS always outputs isotropic voxel size so a resolution of type -r=1.5x1.75x2.5. is not accepted. it should be "-r 1.5" to have the outputs be at 1.5mm isotropic resolution.
2) At the begininng, when reading tensors, all the tensors are named "tortoiset_aff.nii" , this generally indicates a problem with the text file sus.txt. Are all the lines "tortoiset.nii" ?
3) A spacing of 0 is not allowed. The resolution of the input files seem to be weird. Are these tensors, specifically tortoiset.nii computed by TORTOISE? I mean did you use an executable like EstimateTensorWLLS or EstimateTensorNLLS?
The resolution of a tensor image should be of the form resx x resy x resz x 1
The tensors computed by TORTOISE executables would already be in this form.
Let us know.