How to build a CNN to analyze volumetric images?

Question

3D

Hello, the situation is as follows: there are flat images from which you can make up a three-dimensional image (as far as I know, images are also stored, for example, in MRI). So, for their analysis it is planned to write a grid, but I have no idea how to work with volume.
Does anyone have any experience?
Thanks

Answer 1

you are confusing stereophotogrammetry with post-MRT DICOM format, and it is not clear what processing you are going to use.
I don’t know about libraries, but there are ready-made free solutions for both options.

Answer 2

The task is not entirely clear. If it consists in 3D reconstruction from 2D images, then usually more direct mathematical methods are used: characteristic points are selected on all images, and then the points of one image correspond to the points of other images. And as a final stage, 3D reconstruction.
If the task is simply to select some parts on an already finished 3D (for example, if 2D images are slices in parallel planes and reconstruction is essentially not required), then you can try to adapt convolutional networks for such a task, but most likely you will have to write all by yourself - ready-made libraries for 3D hardly exist.