Page MenuHomePhabricator

Segmentation created for multiple timepoints on 3D+t images
Closed, ResolvedPublic

Description

If you create a segmentation for e.g. the first timestep of a 3D+t image, the segmentation is automatically created for the last timepoint as well. This does not make sense.

Reproduce:

  • image: MITK-Data/3D+t-Heart
  • create a new segmentation on the first timestep
  • check the last timestep

Event Timeline

kausch triaged this task as Normal priority.Apr 12 2019, 3:59 PM
kausch created this task.

I couldn't reproduce it. @kleina Can you? (maybe it occurred only due to missing reinit?)

I can reproduce it. I did exactly the steps in the task description. I used MITK.2018.4.0 installer. I'll check on a (more or less) current master.

I forgot to enable the segmentation plugin... I'll test it tomorrow.

Ok, I tested it on my current developing build, which is branched off of master from 2019-04-29 (a0ba9d263118bc2df2aa922c124c3db57acbd2dc). I can reproduce the error as described in the description. So I guess to answer your question @kislinsk: The status is "still open".

I just tried this and I can reproduce it with MITK-Data/3D+t-Heart. But when I used Pic2DplusT the segmentation was only created on a single timestep.
Additionally, when I clicked on the arrows of the time-slider of the image navigation plugin the time slider sometimes wraps around the last time step and returns to timestep 0. This was already mentioned here T25030. However, this does not happen If I use the Pic2DplusT image so maybe the 3D+t-Heart image is corrupt?

I also tried it with MITK-Data/3D+t-ITKIO-TestData and everything works fine (segmentation and time-slider). However, a difference between all these time-images and the 3D+t-Heart// image is, that the Heart-image does not actually have a third timepoint, as the third timepoint is of length 0ms.
So maybe there is a problem with the matching of time points and the segmentation is displayed for the last time point, as the lats time point is actually non-existent and wraps around to the time point 0.

I created such a segmentation on the 3D+t-heart dataset and stored the segmentation as *.nrrd. The I loaded the *.nrrd-segmentation again and I now I can only see the segmentation on the first timestep.
Important: The timesteps of the *.nrrd-segmentation are now from 0 - 3 and with a duration of 1ms. The original 3D+t-Heart was [0,103] [103,160] [160, 160]. This was also true for the original segmentation. If I store the 3D+t-Heart image as *.nrrd and reopen it again the original timesteps are preserved.

Will be solved if D214 is accepted and merged.

kalali claimed this task.