Improved Complex Convolutional Neural Network Based on SPIRiT and Dense Connection for Parallel MRI Reconstruction

Table 3

Comparisons of average PSNR and SSIM values of the MR images reconstructed by all the models for the Coronal-PD dataset from the 3x and 5x accelerated 1DRU, 2DRU, RADU, and 2DPU patterns.

Undersampling; patterns	Reconstruction; models	Threefold AF		Fivefold AF
		PSNR (dB)	SSIM	PSNR (dB)	SSIM
1DRU	SPIRiT	31.970 ± 1.157	0.823 ± 0.028	29.261 ± 1.683	0.761 ± 0.039
	Deepcomplex	35.586 ± 1.385	0.880 ± 0.021	33.039 ± 1.243	0.818 ± 0.029
	DONet	35.812 ± 1.443	0.884 ± 0.020	33.540 ± 1.148	0.831 ± 0.028
	SPIRiT-Net	36.468 ± 1.141	0.892 ± 0.019	34.077 ± 1.141	0.843 ± 0.026
2DRU	SPIRiT	38.334 ± 2.279	0.910 ± 0.023	35.782 ± 2.185	0.864 ± 0.032
	Deepcomplex	39.632 ± 2.817	0.931 ± 0.024	37.245 ± 2.684	0.889 ± 0.036
	DONet	39.823 ± 2.984	0.932 ± 0.025	37.500 ± 2.918	0.892 ± 0.038
	SPIRiT-Net	40.937 ± 1.984	0.943 ± 0.016	38.477 ± 1.688	0.910 ± 0.020
2DPU	SPIRiT	39.398 ± 2.762	0.928 ± 0.026	36.832 ± 2.450	0.883 ± 0.034
	Deepcomplex	39.491 ± 2.704	0.934 ± 0.022	37.151 ± 2.450	0.894 ± 0.034
	DONet	39.786 ± 2.846	0.936 ± 0.025	37.597 ± 2.791	0.898 ± 0.037
	SPIRiT-Net	40.869 ± 1.702	0.942 ± 0.016	38.941 ± 1.606	0.919 ± 0.019
RADU	SPIRiT	38.357 ± 2.409	0.911 ± 0.025	35.829 ± 2.533	0.859 ± 0.038
	Deepcomplex	39.803 ± 2.886	0.933 ± 0.025	36.947 ± 2.733	0.883 ± 0.040
	DONet	39.857 ± 3.004	0.933 ± 0.025	37.090 ± 2.847	0.884 ± 0.040
	SPIRiT-Net	40.963 ± 1.847	0.943 ± 0.016	38.011 ± 1.813	0.901 ± 0.023

The best values are in bold.