search

sct-pipeline / contrast-agnostic-softseg-spinalcord

Contrast-agnostic spinal cord segmentation project with softseg
MIT License
4 stars 3 forks source link

Loss of available segmentation if FOV of T2w is smaller than T1w #12

Open sandrinebedard opened 3 years ago

sandrinebedard commented 3 years ago

Description

The current pipeline to generate softseg aligns all images to T2w space. However, there are some cases that the FOV of T1w in S-I direction is larger than for T2w. In those cases, the segmentation of T1w out of bounds for T2w space is lost during registration to T2w and when we bring back the softseg, it ends before the end of spinal cord (and original segmentation)

Original segmentation for T1w

image

Softseg brought back from T2w space to T1w:

image

Not sure if we really want to address this, I haven't investigated yet how many images this affects. If we want to address this, it could be possible to add the missing part of the segmentation since we can know the minimum index of the softseg in T1w space.

jcohenadad commented 3 years ago

I know this would happen eventually 😅 For now, let's not consider these missing segmentations. I know this is frustrating, but we need to move forward and this is not top priority. We can leave the issue open and address it at some point.

sandrinebedard commented 3 years ago

I know this would happen eventually 😅 For now, let's not consider these missing segmentations. I know this is frustrating, but we need to move forward and this is not top priority. We can leave the issue open and address it at some point.

I totally agree!