Application of Compressive Sensing to Ultrasound Images: A Review

Table 1

CS-based reconstruction algorithms.

S/N	References	Method	NMSE after 30 iterations				SSIM
1	[32]	Approximate messaging passing DCT, wavelet, and spatial as transform domain; Soft thresholding and ABE as denoiser	Time soft thresholding			−5 (dB)
			Time ABE			−5 (dB)
			Wavelet soft thresholding			−10 (dB)
			Wavelet ABE			−10.12 (dB)
			Discrete cosine transform ST			−13.97 (dB)
			Discrete cosine transform ABE			−21.23 (dB)
2	[33]	Approximate messaging passing Cauchy prior-based maximum a posteriori	Algorithm		Time (sec)
			ST		4.57	−14.25 (dB)
			ABE		4.77	−15.15 (dB)
			Cauchy-MAP		5.33	−16.27 (dB)
3	[34]	IRLS-DP; FD-SαS-IRLS; SαS-IRLS; LP minimization; (DP) dual prior information	NRMSE				Sαs-IRLS	FD-SαSIRLS	IRLS-DP
			Compression ratio	Sαs-IRLS	FD-SαSIRLS	IRLS-DP
			33%	0.697	0.540	0.249	0.208	0.586	0.908
			50%	0.518	0.291	0158	0.377	0.844	0.944
4	[35]	CS-STA; Sym8 wavelet as sparsifier	NRMSE
			CS-STA	32	64	128
			Results 1	0.98%	0.42%	0.01%
			Results 2	0.41%	0.12%	0.001%
5	[36]	Bayesian framework-based algorithm; Fourier transform as sparse domain	NRMSE
				Simulated image		In vivo image
			Classical CS	E = 0.12		E = 0.10
			Bayesian CS	E = 0.12		E = 0.07