Partial Volumes Generated do not spatially align with VolumeNode Input

[amymmorton]

As seen with Cesar's data .. I tried a 3d static ct acquired by Dr Lalon'es group:

spatial mismatch of generated tfm/tiffs:

nrrd and seg

[amymmorton]

for ref: this is a bilateral hand/wrist ct scan, proper seg:

[amymmorton]

edit - trying to define exact cases where partial volumes (tiff) and tfm are correct and when they are not 'example fail of previously processed (working) data set:

[jcfr]

Thanks for feedback.

To support you with the demo, we provides you with extension packages for the past few version, that way you could install it locally and use the version that works. Let us know if that would help.

Looks like our testing didn't account for this dataset and/or usecase.

We should work together to identify which changes introduced the regression.

For reference, here are the recent SlicerAutoscoperM changes:

[amymmorton]

IN regards to this data

loaded from nrrd and seg

IN slicer/slicer@c6b01db correct PV formed.

sourcebuild lib zip in Gdrive link as well

scmrevision 2e483f4

[amymmorton]

dataset (sid)	segmentation type	link	align screenshot
WRIST BVR (WN00112)	STL Model	link
THUMB 3DCT (BN00124)	STL Model	link
KNEE BVR (CXBM15)	STL Model	link
BILATERL HAND CT	STL Model	link

[sbelsk]

Reopening as #128 didn't resolve the problem, but fixed a related bug.

[amymmorton]

Tue Nov 12 2024 (Revision: 34fa24c) WRIST BVR (WN00112) link

Al PV correct EXCEPT the cropped volume node use case(last)

segmentation type	align screenshot
STL Model
Auto Seg
Auto Seg (from crop roi, Partial V original neutral)
Auto Seg (from crop roi, Partial V CROP neutral)

[amymmorton]

Tue Nov 12 2024 (Revision: https://github.com/BrownBiomechanics/SlicerAutoscoperM/commit/34fa24cd66f0b3402a1e8ea09cacef4b862e40a1)

BILATERL HAND CT

Import to dicom db from file (rather than previously saved .nrrd)

segmentation type	align screenshot
auto seg

[amymmorton]

@jcfr

Could the issue be with the IJK - RAS (LPS) differences:

WRIST: (successful)	BILATERAL: (fail)

[amymmorton]

MUST BE!

When the only diff between original and cropped as well!

[sbelsk]

Excellent find, thanks for looking into it! I opened https://github.com/BrownBiomechanics/SlicerAutoscoperM/pull/147, which should fix the problem.