Fig. 1: An x-ray nanoprobe illuminates an IC sample at different angles, where multi-slice model is employed to account for the diffraction effect.
From: Deep learning enables nanoscale X-ray 3D imaging with limited data

With limited number of projections, a poor-quality approximant was first obtained using ptycho-tomography algorithm, and then fed to the 3D U-net to produce a nearly-perfect tomogram. The network was trained on the reconstruction with the full dataset acquired from the same sample