notesum.ai

Published at November 27

Deep End-to-end Adaptive k-Space Sampling, Reconstruction, and Registration for Dynamic MRI

eess.IV

cs.CV

physics.med-ph

Released Date: November 27, 2024

Authors: George Yiasemis¹, Jan-Jakob Sonke¹, Jonas Teuwen²

Aff.: ¹Netherlands Cancer Institute, Amsterdam, Netherlands; University of Amsterdam, Amsterdam, Netherlands; ²Netherlands Cancer Institute, Amsterdam, Netherlands; University of Amsterdam, Amsterdam, Netherlands; Radboud University Medical Center, Nijmegen, Netherlands

Arxiv: http://arxiv.org/abs/2411.18249v1

Reconstruction Model	Sampling Type	Sampling Initialization	Registration Module	Joint Loss	Loss Weights		$4\times$			$6\times$			$8\times$
Reconstruction Model	Sampling Type	Sampling Initialization	Registration Module	Joint Loss	$\alpha$	$\beta$	SSIM ( $\uparrow$ )	PSNR ( $\uparrow$ )	NMSE ( $\downarrow$ )	SSIM ( $\uparrow$ )	PSNR ( $\uparrow$ )	NMSE ( $\downarrow$ )	SSIM ( $\uparrow$ )	PSNR ( $\uparrow$ )	NMSE ( $\downarrow$ )
vSHARP	Adaptive	✗	Learned	✓	1	1	0.877 $\pm$ 0.050	31.68 $\pm$ 2.86	0.037 $\pm$ 0.019	0.867 $\pm$ 0.052	31.41 $\pm$ 2.76	0.038 $\pm$ 0.019	0.854 $\pm$ 0.054	31.01 $\pm$ 2.69	0.042 $\pm$ 0.020
vSHARP	Adaptive	✗	Learned	✗	1	1	0.880 $\pm$ 0.048	31.95 $\pm$ 2.81	0.034 $\pm$ 0.018	0.872 $\pm$ 0.049	31.75 $\pm$ 2.70	0.035 $\pm$ 0.017	0.862 $\pm$ 0.052	31.41 $\pm$ 2.64	0.038 $\pm$ 0.018
vSHARP	Adaptive	✓	Learned	✓	1	1	0.882 $\pm$ 0.049	31.90 $\pm$ 2.92	0.035 $\pm$ 0.019	0.874 $\pm$ 0.051	31.75 $\pm$ 2.84	0.036 $\pm$ 0.019	0.862 $\pm$ 0.054	31.40 $\pm$ 2.75	0.039 $\pm$ 0.020
vSHARP	Adaptive	✓	Learned	✗	1	1	0.881 $\pm$ 0.049	31.92 $\pm$ 2.91	0.035 $\pm$ 0.019	0.873 $\pm$ 0.051	31.73 $\pm$ 2.79	0.036 $\pm$ 0.018	0.861 $\pm$ 0.054	31.35 $\pm$ 2.75	0.039 $\pm$ 0.020
vSHARP	Equi	NA	Learned	✓	1	1	0.879 $\pm$ 0.050	31.62 $\pm$ 2.98	0.038 $\pm$ 0.021	0.868 $\pm$ 0.053	31.51 $\pm$ 2.88	0.038 $\pm$ 0.020	0.848 $\pm$ 0.057	30.93 $\pm$ 2.72	0.043 $\pm$ 0.021
vSHARP	Equi	NA	Learned	✗	1	1	0.881 $\pm$ 0.050	31.75 $\pm$ 3.00	0.037 $\pm$ 0.020	0.869 $\pm$ 0.054	31.62 $\pm$ 2.91	0.037 $\pm$ 0.020	0.849 $\pm$ 0.057	31.05 $\pm$ 2.74	0.042 $\pm$ 0.020
vSHARP	Optimized	✗	Learned	✓	1	1	0.886 $\pm$ 0.047	32.17 $\pm$ 2.87	0.033 $\pm$ 0.018	0.879 $\pm$ 0.048	31.96 $\pm$ 2.80	0.034 $\pm$ 0.018	0.870 $\pm$ 0.050	31.71 $\pm$ 2.72	0.036 $\pm$ 0.018
VarNet	Adaptive	✗	Learned	✓	1	1	0.837 $\pm$ 0.056	30.92 $\pm$ 2.40	0.041 $\pm$ 0.015	0.832 $\pm$ 0.057	30.77 $\pm$ 2.40	0.043 $\pm$ 0.016	0.829 $\pm$ 0.058	30.66 $\pm$ 2.38	0.044 $\pm$ 0.016
vSHARP	Adaptive	✗	OFILK	NA	1	NA	0.868 $\pm$ 0.049	31.53 $\pm$ 2.68	0.037 $\pm$ 0.019	0.856 $\pm$ 0.052	31.15 $\pm$ 2.65	0.041 $\pm$ 0.019	0.842 $\pm$ 0.056	30.70 $\pm$ 2.62	0.045 $\pm$ 0.021
vSHARP	Adaptive	✗	OFTVL1	NA	1	NA	0.873 $\pm$ 0.045	30.01 $\pm$ 2.78	0.055 $\pm$ 0.031	0.861 $\pm$ 0.049	29.55 $\pm$ 2.76	0.061 $\pm$ 0.034	0.848 $\pm$ 0.053	29.01 $\pm$ 2.77	0.069 $\pm$ 0.038
vSHARP	Adaptive	✗	DEMONS	NA	1	NA	0.873 $\pm$ 0.050	30.36 $\pm$ 2.80	0.049 $\pm$ 0.024	0.863 $\pm$ 0.053	30.19 $\pm$ 2.78	0.051 $\pm$ 0.025	0.852 $\pm$ 0.056	29.93 $\pm$ 2.73	0.054 $\pm$ 0.026
vSHARP	Equi	NA	OFILK	NA	1	NA	0.877 $\pm$ 0.046	31.82 $\pm$ 2.74	0.035 $\pm$ 0.018	0.861 $\pm$ 0.050	31.43 $\pm$ 2.65	0.038 $\pm$ 0.018	0.838 $\pm$ 0.055	30.72 $\pm$ 2.50	0.044 $\pm$ 0.019
vSHARP	Equi	NA	OFTVL1	NA	1	NA	0.882 $\pm$ 0.044	30.40 $\pm$ 2.99	0.051 $\pm$ 0.031	0.868 $\pm$ 0.047	30.06 $\pm$ 2.87	0.054 $\pm$ 0.031	0.847 $\pm$ 0.052	29.32 $\pm$ 2.72	0.064 $\pm$ 0.035
vSHARP	Equi	NA	DEMONS	NA	1	NA	0.880 $\pm$ 0.048	30.55 $\pm$ 2.83	0.047 $\pm$ 0.024	0.865 $\pm$ 0.052	30.32 $\pm$ 2.77	0.050 $\pm$ 0.024	0.845 $\pm$ 0.056	29.89 $\pm$ 2.64	0.054 $\pm$ 0.025