Skip to main content
. Author manuscript; available in PMC: 2023 May 11.
Published in final edited form as: Magn Reson Med. 2021 Jul 13;86(5):2666–2683. doi: 10.1002/mrm.28912

Table 1.

Quantitative evaluation: SSIM3D and nRMSE are calculated on reconstructed results from all patients (N=10) in test dataset B1 and mean and standard deviation (Std. Deviation) of them over the patients are reported for different methods. Based on the multiple pair comparisons, there is a statistically significant difference (P<0.05) between the SSIM and nRMSE metrics of the 2D-GAN reconstruction images and other methods. The proposed method (TAV-GAN) achieved the highest SSIM and the lowest nRMSE among the other methods.

Methods SSIM nRMSE
Mean Std. Deviation Mean Std. Deviation
ZF 0.376S1 0.0446 0.094S1 0.0194
2D-GAN 0.481S2 0.0594 0.072S2 0.0138
3D U-Net 0.732 0.0483 0.040 0.0085
Volumetric-GAN 0.752 0.0479 0.038 0.0090
Temporal-GAN 0.746 0.0495 0.036 0.0072
TAV-GAN 0.785 0.0389 0.030 0.0058
S1

There was a statistically significant difference (P<0.05) between the ZF method and other methods with respect to the quantitative metrics SSIM and nRMSE.

S2

There was a statistically significant difference (P<0.05) between the 2D-GAN method and other methods with respect to the quantitative metrics SSIM and nRMSE.