How to use SaveImaged

[JumpLK]

Hi, I'm using the data from the medical segmentation decathlon(the liver ones) : http://medicaldecathlon.com
I'm trying to use SaveImaged to see how my transform affects the data. I tried multiple things :

1. To save images after my transform :
   
   It seems like by doing so it tries to allocate a lot of memory (735 Gb):
   

2. To save images just after the load :
   
   It produces an output but it looks corrupted in some way because I'm not able to read it with Slicer3D it makes the software crash

3. To save only the image
   
   Same as 2. : It produces an output but it looks corrupted in some way because I'm not able to read it with Slicer3D it makes the software crash

I think that all of thoose errors are because I'm not using SaveImaged as it's intendeed to be used, so do you know how should I use it ?

[rijobro]

Hi @JumpLK. SaveImage and SaveImaged require the image to be passed with the channel dimension first. So firstly you should add AddChanneld just after the LoadImaged. This also might solve your data allocation problem (currently Spacingd and Orientationd will be assuming that you have an image that is 512x75 with 512 channels).

I took your code, added the AddChanneld and then could open the saved image in ITKSnap. However, image metadata still says that the image is 4D (with a singleton channel dimension). This may not cause any problems for you, but I've submitted a PR to allow the user to squeeze the final dimensions, which would remove the singleton channel dimension, giving a 3D nifti image again.

[JumpLK]

Thank you, it works :)