Undefined appearance of the spinal cord in the generated template

[jcohenadad]

As per #16, the spinal cord is defined on the template as being much smaller than the area that seems to coincide with the CSF boundary:

I suspect this is a glitch that was produced when non-linearly registering and then averaging several dog images.

To rule in/out this hypotheses, this issue looks at individual dog images around this lumbar region.

[jcohenadad]

sub-HarshmanDobby:

Spinal cord / CSF boundary is very clear, without a 'circle' between the cord and the CSF:

[jcohenadad]

sub-JacobsenPetri:

As for the previous dog, the boundary is clear:

[rohanbanerjee]

sub-JamesMonk:

The dark circle around the spinal cord is visible in this subject.

[rohanbanerjee]

sub-TrainqueNewt:

The circle is not visible around the spinal cord in this subject.

[rohanbanerjee]

sub-WalkerBo:

The dark circle around the spinal cord is visible in this subject.

[jcohenadad]

ALWAYS tell us in the GH issue what your plan is BEFORE moving forward with analysis (eg: i don’t agree with removing one subject– it does not make sense)-- also, i do NOT agree that one dog shows the ‘dark circle’. Also. These images only show one axial slice– this is not a thorough overview. I think that all dogs should be considered for this template. In some cases, we do see a dark ring, but I don't think this will be a problem when generating the average template (it will average out).

A more important think that needs to be done, is to investigate how each individual dog’s spinal cord look across the template generation pipeline (eg: after 5 iterations: how does it look like? is there a dog where the registration is completely ‘off’, which would explain the weird appearance in the final template?). If that's the case, then what needs to be done to fix it is to provide manual segmentation for individual dogs during template generation process.

[rohanbanerjee]

Referred to the (a) MNI-AMU-Poly template paper, (b) PAM50 paper, (c) Unbiased average age-appropriate atlases for pediatric studies (in which the template generation pipeline we are using was introduced) to understand the effect of parameters.

I drew inspiration from the above manuscripts to change parameters mentioned in this part of the template generation pipeline. Below are the trials:

1. Tried to change the number of a iterations with multiple variations. This was to test if the presence of the surrounding circle fades away at a lower resolution:
   --> 4 iterations at 4mm resolution, 4 iterations at 2mm resolution, 4 iterations at 1mm resolution, 4 iterations at 0.5mm resolution. Template after 6 iterations can be found in the attached folder with the name: avg006_4_2.nii
   --> 4 iterations at 16mm resolution, 4 iterations at 8mm resolution, 4 iterations at 4mm resolution, 4 iterations in 2mm resolution, 8 iterations in 1mm resolution. Template after 8 iterations can be found in the attached folder with the name: avg008_16_8.nii

The dark circle around the spinal cord still exists.

2. Checking the individual dog's spinal cord after 16 iterations during the template generation. The dark circle is still visible for all the dogs. Individual dog can be found in the attached folder
   (Naming convention --> HarshmanDobby: avg016_16_8_HD_norm.nii, JacobsenPetri: avg016_16_8_JP_norm.nii, JamesMonk: avg016_16_8_JM_norm.nii, TrainqueNewt: avg016_16_8_TN_norm.nii, WalkerBo: avg016_16_8_WB_norm.nii)

Link to outputs from trials: https://drive.google.com/file/d/1dPfYtxXxMrg66cA0yxVGhww8C68zBP98/view?usp=share_link

[jcohenadad]

avg016_16_8_TN_norm.nii and avg016_16_8_WB_norm.nii are the same dog