Figure 5.
The diagram presents three parallel workflows from an input sinogram. In the top row, interpolation produces a linearly interpolated sinogram, which is processed by a Neural Network to generate a deep learning synthesised sinogram, labelled as Sinogram domain preprocessing. In the middle row, F B P reconstructs an incomplete C T image, which is refined by a Neural Network to obtain a deep learning C T image, labelled as C T image domain post-processing. In the bottom row, a Neural Network first generates a full view sinogram, followed by F B P to reconstruct a full view C T image, and a final Neural Network produces a deep learning C T image, labelled as Dual domain joint optimisation.Image domain processing methods