Generate an approximate spinal cord mask for the phantom data

spinal-cord-7t / coil-qc-code

7T Spinal Cord Coil QC Analysis Code

0 stars 0 forks source link

Generate an approximate spinal cord mask for the phantom data #71

Closed jcohenadad closed 3 weeks ago

jcohenadad commented 1 month ago

Proposed strategy:

Identify a reference site
Identify a reference image
Generate an approximate mask of the spinal cord on the phantom for this site
Register the reference image from site i to the reference site https://github.com/spinal-cord-7t/coil-qc-code/issues/72
Apply the inverse transformation to the mask (using trilinear interpolation).

Related to #63

jcohenadad commented 1 month ago

One option would be to use the anat/sub-CRMBM_T2starw, but given the presence of gradient non-linearity distortions (see below), corrected on the sub-CRMBM_T2starw but not corrected on the fmap data, which are ultimately used for quantitative analysis, I think we should instead rely on the sub-CRMBM_acq-anat_TB1TFL to generate the mask.

ezgif-4-3851abf0d3

jcohenadad commented 1 month ago

Problem: sub-MSSM site does not have a _acq-anat_TB1TFL file.

jcohenadad commented 1 month ago

Looking at the variability of phantom scanning across sites to evaluate strategy for co-registration:

ezgif-4-e9f1a8aa4c

Notes:

Strong bias field (might want to try N4)
Mean image intensity varies across sites-- thresholding method won't work
Slight image distortion towards the posterior edge, eg MPI

Possible strategies

Mask-based registration method, eg with a mask covering the posterior edge of the phantom
Use bias field correction
Manual marker-based registration approach (would account for sub-MSSM site)

jcohenadad commented 1 month ago

For manual marker, we could use the posterior tip of the neck, at the inflexion point: