. 2021 May 15;23:100271. doi: 10.1016/j.pacs.2021.100271

Table 3.

Comparison of reconstruction performance for mouse cerebral vasculature in different models trained with either textural images or both the textural images and measured time-series signals.

Dataset	Mouse Cerebral Vasculature
Network	Parameters	PSNR	SSIM	SSIM Luminance	SSIM Contrast	SSIM Structure	Exec.
FD-UNet (TR)	37,888,449	19.779 ± 1.116	0.599 ± 0.028	0.838 ± 0.039	0.907 ± 0.016	0.744 ± 0.044	42 ms
FD-UNet (BP)	37,888,449	20.808 ± 1.189	0.643 ± 0.025	0.863 ± 0.036	0.926 ± 0.012	0.771 ± 0.040	42 ms
Y-Net (TR & Raw)	37,506,354	19.000 ± 1.545	0.623 ± 0.048	0.836 ± 0.029	0.900 ± 0.016	0.766 ± 0.046	48 ms
Y-Net (BP & Raw)	37,506,354	20.973 ± 1.246	0.667 ± 0.043	0.857 ± 0.029	0.922 ± 0.014	0.796 ± 0.041	48 ms
FD-YNet (TR & Raw)	34,852,306	19.580 ± 1.376	0.621 ± 0.052	0.818 ± 0.036	0.896 ± 0.020	0.776 ± 0.044	70 ms
FD-YNet (BP & Raw)	34,852,306	21.222 ± 1.346	0.674 ± 0.043	0.867 ± 0.025	0.925 ± 0.014	0.796 ± 0.041	70 ms

The abbreviations in parentheses represent the data form used by the models. TR: time-reversal image. BP: backprojection image. Raw: measured time-series signals. PI: pixel-interpolated data.