Table 4 Quantitative analysis of image quality and performance of deep-learning models in the reconstructed images from different methods with dataset No. 1 in Group 2
From: Artifact suppression for breast specimen imaging in micro CBCT using deep learning
| Â | SSIM | CNR | PSNR |
|---|---|---|---|
Image from FBP | 0.025 | 0.960 | 6.510 |
Image from IR | 0.382 | 3.180 | 15.292 |
Image from the sinogram with linear interpolation | 0.788 | 7.100 | 28.055 |
Image from denoise with ResU-Net | 0.816 | 7.710 | 28.020 |
Image from the sinogram with U-Net | 0.854 | 8.010 | 22.541 |
Image from the synthesized sinogram with the modified U-Net and ResU-Net models in the proposed method | 0.866 | 8.380 | 28.726 |