Table 4 Zero-shot learning evaluation metrics for equispaced Cartesian under-sampling, 1D Gaussian under-sampling, and 2D Gaussian under-sampling reconstructions.
From: Learning to reconstruct accelerated MRI through K-space cold diffusion without noise
| Â | 4x | 8x | 4x with 8-fold model | |||
|---|---|---|---|---|---|---|
PSNR | SSIM | PSNR | SSIM | PSNR | SSIM | |
Equispaced Cartesian Under-sampling | ||||||
 U-Net | 28.56 | 0.6201 | 27.07 | 0.5406 | 27.53 | 0.6067 |
 W-Net | 29.63 | 0.6730 | 28.34 | 0.6109 | 28.56 | 0.6591 |
 E2E-VarNet | 30.25 | 0.6821 | 29.10 | 0.6210 | 29.53 | 0.6583 |
 K-space cold diffusion | 30.54 | 0.7139 | 29.56 | 0.6440 | 30.50 | 0.7138 |
1D Gaussian Under-sampling | ||||||
 U-Net | 28.70 | 0.6311 | 27.21 | 0.5579 | 27.53 | 0.6258 |
 W-Net | 29.61 | 0.6780 | 28.50 | 0.6253 | 28.32 | 0.6778 |
 E2E-VarNet | 30.19 | 0.6879 | 29.37 | 0.6309 | 29.47 | 0.6617 |
 K-space cold diffusion | 30.34 | 0.7090 | 29.46 | 0.6464 | 30.17 | 0.7061 |
2D Gaussian Under-sampling | ||||||
 U-Net | 27.90 | 0.6019 | 26.03 | 0.5193 | 26.28 | 0.5951 |
 W-Net | 28.97 | 0.6671 | 27.49 | 0.5989 | 27.45 | 0.6653 |
 E2E-VarNet | 28.30 | 0.6226 | 21.83 | 0.4067 | 20.82 | 0.4286 |
 K-space cold diffusion | 30.04 | 0.6992 | 29.31 | 0.6341 | 30.04 | 0.6999 |
