Parallel Imaging Reconstruction on knee dataset CORPD-weighted in-distribution (test)
37.05PSNRfully-learned filters and time conditioning (FLF+TC)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| fully-learned filters and time conditioning (FLF+TC)k-space Trajectory=Cartesian, Acceleration=8, Auto Calibration Lines=8 (Rotated phase-encoding direction), Number of Parameters=73482026.05 | 37.05 | 92 | 0.24 | |
| end-to-end variational network (VN)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=8, Number of Parameters=3 × 10^72026.05 | 36.92 | 92 | 0.24 | |
| fully-learned filters (FLF)k-space Trajectory=Cartesian, Acceleration=8, Auto Calibration Lines=8 (Rotated phase-encoding direction), Number of Parameters=17602026.05 | 36.68 | 91 | 0.26 | |
| Shearlet (SH)k-space Trajectory=Cartesian, Acceleration=8, Auto Calibration Lines=8 (Rotated phase-encoding direction), Number of Parameters=15582026.05 | 36.19 | 91 | 0.29 | |
| fully-learned filters and time conditioning (FLF+TC)k-space Trajectory=2D Gaussian, Acceleration=8, Number of Parameters=73482026.05 | 35.84 | 89 | 0.32 | |
| fully-learned filters (FLF)k-space Trajectory=2D Gaussian, Acceleration=8, Number of Parameters=17602026.05 | 35.75 | 89 | 0.32 | |
| Shearlet (SH)k-space Trajectory=2D Gaussian, Acceleration=8, Number of Parameters=15582026.05 | 35.6 | 89 | 0.33 | |
| fully-learned filters and time conditioning (FLF+TC)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=8, Number of Parameters=73482026.05 | 35.5 | 89 | 0.36 | |
| fully-learned filters (FLF)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=8, Number of Parameters=17602026.05 | 35.39 | 89 | 0.36 | |
| fully-learned filters and time conditioning (FLF+TC)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=4, Number of Parameters=73482026.05 | 35.32 | 89 | 0.36 | |
| fully-learned filters (FLF)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=4, Number of Parameters=17602026.05 | 35.29 | 89 | 0.36 | |
| fully-learned filters and time conditioning (FLF+TC)k-space Trajectory=Radial, Acceleration=11, Number of Parameters=73482026.05 | 35.1 | 87 | 0.38 | |
| fully-learned filters (FLF)k-space Trajectory=Radial, Acceleration=11, Number of Parameters=17602026.05 | 34.94 | 87 | 0.38 | |
| Shearlet (SH)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=8, Number of Parameters=15582026.05 | 34.64 | 88 | 0.41 | |
| Shearlet (SH)k-space Trajectory=Radial, Acceleration=11, Number of Parameters=15582026.05 | 34.54 | 87 | 0.42 | |
| Deep-JSense (DJS)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=8, Number of Parameters=4474902026.05 | 34.45 | 90 | 0.46 | |
| total variation (TV)k-space Trajectory=2D Gaussian, Acceleration=82026.05 | 34.35 | 86 | 0.44 | |
| Shearlet (SH)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=4, Number of Parameters=15582026.05 | 34.3 | 88 | 0.44 | |
| Deep-JSense (DJS)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=4, Number of Parameters=4474902026.05 | 33.84 | 90 | 0.53 | |
| total variation (TV)k-space Trajectory=Cartesian, Acceleration=8, Auto Calibration Lines=8 (Rotated phase-encoding direction)2026.05 | 33.65 | 84 | 0.51 | |
| total variation (TV)k-space Trajectory=Radial, Acceleration=112026.05 | 33.24 | 82 | 0.56 | |
| Deep-JSense (DJS)k-space Trajectory=Cartesian, Acceleration=8, Auto Calibration Lines=8 (Rotated phase-encoding direction), Number of Parameters=4474902026.05 | 33.2 | 90 | 1.85 | |
| total variation (TV)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=82026.05 | 33 | 83 | 0.59 | |
| Deep-JSense (DJS)k-space Trajectory=Radial, Acceleration=11, Number of Parameters=4474902026.05 | 32.79 | 86 | 0.71 | |
| end-to-end variational network (VN)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=4, Number of Parameters=3 × 10^72026.05 | 32.16 | 89 | 0.69 | |
| Zero filled (ZF)k-space Trajectory=2D Gaussian, Acceleration=82026.05 | 32.15 | 84 | 0.78 | |
| total variation (TV)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=42026.05 | 31.86 | 82 | 0.81 | |
| Zero filled (ZF)k-space Trajectory=Cartesian, Acceleration=8, Auto Calibration Lines=8 (Rotated phase-encoding direction)2026.05 | 31.13 | 81 | 0.97 | |
| Zero filled (ZF)k-space Trajectory=Radial, Acceleration=112026.05 | 28.76 | 75 | 1.67 | |
| Zero filled (ZF)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=82026.05 | 27.19 | 74 | 2.37 | |
| Deep-JSense (DJS)k-space Trajectory=2D Gaussian, Acceleration=8, Number of Parameters=4474902026.05 | 25.44 | 80 | 6.63 | |
| end-to-end variational network (VN)k-space Trajectory=Cartesian, Acceleration=8, Auto Calibration Lines=8 (Rotated phase-encoding direction), Number of Parameters=3 × 10^72026.05 | 24.72 | 67 | 3.29 | |
| Zero filled (ZF)k-space Trajectory=Cartesian, Acceleration=4, Auto Calibration Lines=42026.05 | 24.14 | 69 | 4.96 | |
| end-to-end variational network (VN)k-space Trajectory=2D Gaussian, Acceleration=8, Number of Parameters=3 × 10^72026.05 | 23.53 | 73 | 4.33 | |
| end-to-end variational network (VN)k-space Trajectory=Radial, Acceleration=11, Number of Parameters=3 × 10^72026.05 | 20.56 | 69 | 6.55 |