Table 3.
Evaluation of model performance on the mouse brain vasculature in psnr metrics at different noise levels.
| MBLr | FIRe | 2S-FIRe | TV | U-Net | |
|---|---|---|---|---|---|
| 6 dB SNR | 25.568 ± 1.999 | 25.236 ± 2.012 | 25.012 ± 1.831 | 23.204 ± 1.762 | 20.854 ± 1.599 |
| 9 dB SNR | 27.180 ± 2.065 | 26.624 ± 1.998 | 26.170 ± 1.723 | 24.552 ± 1.791 | 21.724 ± 1.619 |
| 12 dB SNR | 28.243 ± 2.091 | 27.614 ± 1.963 | 26.970 ± 1.683 | 25.483 ± 1.789 | 22.184 ± 1.606 |
| 15 dB SNR | 28.695 ± 2.090 | 28.037 ± 1.971 | 27.850 ± 1.638 | 25.911 ± 1.810 | 22.562 ± 1.593 |
| 18 dB SNR | 29.211 ± 2.080 | 28.565 ± 1.964 | 27.743 ± 1.645 | 26.468 ± 1.818 | 22.515 ± 1.594 |
| 21 dB SNR | 29.397 ± 2.066 | 28.740 ± 1.938 | 27.895 ± 1.637 | 26.681 ± 1.824 | 22.580 ± 1.591 |
| 24 dB SNR | 29.481 ± 2.050 | 28.834 ± 1.934 | 27.976 ± 1.618 | 26.776 ± 1.825 | 22.602 ± 1.585 |
| 27 dB SNR | 29.528 ± 2.055 | 28.884 ± 1.933 | 28.011 ± 1.627 | 26.830 ± 1.823 | 22.615 ± 1.587 |
| 30 dB SNR | 29.554 ± 2.048 | 28.911 ± 1.932 | 28.034 ± 1.626 | 26.862 ± 1.822 | 22.624 ± 1.589 |
| - | 29.573 ± 2.054 | 28.918 ± 1.938 | 28.055 ± 1.624 | 26.887 ± 1.824 | 22.628 ± 1.588 |
MBLr: model-based learning. FIRe: fast iterative reconstruction. 2S-FIRe: two-stage fast iterative reconstruction.TV: total variation. U-Net: single-step post-processing. -: no noise included in collected sensor data.