I am trying to build a classifier that would predict the correct outcome (disease vs healthy) using a set of 2D slices derived from the 3D MRI scan. For each patient, based on the 3D scan, I am able to generate the set of slices and then I pick only those that contain region of interest, as a result the number of slices (2D images) vary across the patients - one may have 20 images whereas another 30.
How can I get away with such variable size of the input? Should I be using some RNN as the first part of the architecture where padding/masking could be used? Is anyone aware of some similar works that have been done?
Many thanks, Oleg
I have search for papers where similar architectures would be implemented but couldn't find any