Fig. 7.

Reconstruction results of sinograms of simulation blood vessels using different methods under different projections. (a)-(c) Sparse sinograms (raw data) under 128, 64, and 32 projections, respectively. (d)-(f) Reconstructed sinograms using the Cycle-GAN method. (g)-(i) Reconstructed sinograms using the Attention-based U-Net method. (j)-(l) Reconstructed sinograms using the diffusion model method. (m)-(o) Reconstructed sinograms using the U-Net method. (p)-(r) Reconstructed sinograms using the proposed method. (s) Ground truth sinogram for reference. The yellow numbers at the bottom of each figure are the SSIM values. (t) and (u) The close-up images indicated by the green boxes 1 and 2 in (s), respectively. Yellow arrows highlight differences in reconstructed details between methods. (v) and (w) Signal intensity distributions along the white dashed lines in close-ups 1 and 2 under 32 projections, respectively. (x) and (y) Spectra and phase residuals of the signals along the white dashed lines in close-ups 2, respectively.