notesum.ai

Published at November 19

Enhancing Deep Learning-Driven Multi-Coil MRI Reconstruction via Self-Supervised Denoising

eess.IV

cs.AI

cs.CV

Released Date: November 19, 2024

Authors: Asad Aali¹, Marius Arvinte², Sidharth Kumar³, Yamin I. Arefeen³, Jonathan I. Tamir^3,4,5

Aff.: ¹Department of Radiology, Stanford University; ²Intel Corporation; ³Chandra Family Department of Electrical and Computer Engineering, The University of Texas at Austin; ⁴Dell Medical School, Department of Diagnostic Medicine, The University of Texas at Austin; ⁵Oden Institute for Computational Engineering and Sciences, The University of Texas at Austin

Arxiv: http://arxiv.org/abs/2411.12919v1

Acceleration Factor	Reconstruction Method	Training SNR	Inference SNR
			32dB			22dB			12dB
			NRMSE $\downarrow$	SSIM $\uparrow$	PSNR $\uparrow$	NRMSE $\downarrow$	SSIM $\uparrow$	PSNR $\uparrow$	NRMSE $\downarrow$	SSIM $\uparrow$	PSNR $\uparrow$
4	GSURE-DPS	32dB	0.080	95.43	38.19	0.138	89.91	33.14	0.249	76.36	27.56
		22dB	0.081	95.08	38.03	0.133	91.10	33.36	0.243	78.36	27.74
		12dB	0.078	95.13	38.30	0.126	92.35	33.79	0.233	80.51	28.05
	Naive-DPS	32dB	0.082	96.04	37.99	0.140	88.55	33.07	0.248	73.57	27.70
		22dB	0.085	95.59	37.70	0.145	86.63	32.81	0.262	69.36	27.31
		12dB	0.101	93.35	36.12	0.161	83.37	31.96	0.285	64.88	26.68
	GSURE-MoDL	32dB	0.083	94.99	37.87	0.132	91.79	34.08	0.333	61.18	25.92
		22dB	0.100	92.61	35.79	0.117	92.57	34.78	0.197	86.96	30.49
		12dB	0.141	90.85	32.58	0.146	90.33	32.46	0.169	89.50	31.48
	Naive-MoDL	32dB	0.082	95.89	38.07	0.141	87.58	33.50	0.384	54.98	24.56
		22dB	0.114	91.57	35.06	0.139	87.45	33.41	0.273	65.76	27.48
		12dB	0.183	81.89	30.99	0.192	80.32	30.56	0.256	70.58	27.69
8	GSURE-DPS	32dB	0.164	89.30	30.95	0.214	83.94	28.52	0.297	73.04	25.59
		22dB	0.165	89.22	30.92	0.213	84.59	28.54	0.293	74.76	25.67
		12dB	0.159	89.82	31.22	0.206	86.00	28.80	0.286	76.97	25.85
	Naive-DPS	32dB	0.161	90.18	31.11	0.210	83.76	28.70	0.295	70.81	25.71
		22dB	0.162	89.90	31.05	0.213	82.36	28.62	0.303	67.42	25.55
		12dB	0.176	86.99	30.37	0.226	78.37	28.18	0.325	61.95	25.08
	GSURE-MoDL	32dB	0.132	92.27	33.19	0.170	88.40	31.22	0.348	58.66	25.18
		22dB	0.149	89.64	31.76	0.160	90.20	31.37	0.227	80.42	28.80
		12dB	0.183	86.19	29.84	0.186	86.47	29.79	0.204	86.76	29.29
	Naive-MoDL	32dB	0.133	92.44	33.12	0.177	83.25	30.86	0.368	55.36	24.65
		22dB	0.151	88.81	31.96	0.166	85.93	31.30	0.258	69.38	27.68
		12dB	0.211	78.82	29.03	0.215	78.04	28.94	0.265	70.83	27.11