I have used a 2D GAN code to synthesize CT images from MR.
Since both MR CT comes in 3D, I had to slice the images into 2D images and feed into the network.
Now I want to try the same synthesis project in 3D. I want to use the same GAN algorithm.
How do I shift the same project from 2D to 3D?
What should I be concerned about? How loss, optimization, forward, backward change if data dimension increases?
Can you let me know how did you extracted the slices from the 3d image.I am working on same thing.
Or can you share any reference .
Currently I have my images as 3D , and I am stuck at trying to slice them
I just extracted them axially. Suppose if 3D image dimension is 172x220x156. I created 156 2D images with 156x220 dimension and saved them in .tif format.
It should be pretty straightforward and can be done using Matlab or PIL python library.
Let me know how it goes for you.
Thank you, I was able to do that for 2D network. Have you also tried to create 3D patches.Since I want to create 3D patches for my CT and PET images
Were you successful in converting your 2D project to 3D? As I am also trying to implement the same. But stuck with some errors.
I had converted the 2D synthesis into 3D. If I remember correctly I was having issues in reconstruction from patches.