Please enable JavaScript.
Coggle requires JavaScript to display documents.
Possible solutions, Requirements, Challenges :warning:, Now by solving…
Possible solutions
Graph Neural networks
Ideas
use Christians latent spaces to encode patches as nodes and use similarity between the nodes as edges
-
Why graph NNs
-
Graph (as adjacency matrix) are way less data than high resolution images, yet contains more informative information
-
-
Challenges :warning:
Receptor densities are not of the same resolution ==> it is difficult to have the same patch correspondence between the two modalities
-
-
-
-
-
Now by solving that we will get either a graph for each slice or a graph for the whole brain (it is up to us to decide)
-
-
-
-
-