Temporal super resolution of ultrasound images using compressive sensing. (July 2019)

• Record Type: Journal Article
• Authors: Hosseinpour, Mina; Behnam, Hamid; Shojaeifard, Maryam
• Is Part Of: Biomedical signal processing and control. Volume 52(2019)
• Page Start: 53
• Page End: 68
• Publication Date: 2019-07
• Subjects: Compressing sensing -- Frame rate -- Radio frequency (RF) images sequence -- Spatio-temporal reconstruction -- Sparsity basis -- Intensity variation time curves (IVTC); Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2019.03.003
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: The combination of the temporal and spatial information in compressive sensing (CS) method for frame rate enhancement of 2D and 3D dynamic ultrasound imaging. Using the three sparsity bases including fixed sparsity basis, fixed overcomplete dictionary and learned overcomplete dictionary for sparse representation of the images in Spatio-Temporal domain. Much better quality and less error for reconstructed images using spatio-temporal method than those which reconstructed using conventional spatial CS method. Increasing the frame rate of the ultrasound imaging. Abstract: Increasing the frame rate is a challenging problem for tracking the fast transient motions of the heart in ultrasound imaging with diagnostic goals. In this paper, compressive sensing (CS) is used for super temporal resolution. Compressive sensing is an acquisition method where only a few random samples of a signal are blindly measured, and the full signal is reconstructed under certain conditions. The proposed method uses spatial and temporal information of radio frequency (RF) signals for reconstruction of the entire image sequence so; the reconstruction is performed in both spatial and temporal directions. Three sparsity bases are used for the sparse representation of the signals in the Spatio-Temporal domain, including fixed sparsity basis, fixed overcomplete dictionary and learned overcomplete dictionary. This approach is evaluated on the In-vivo 2-dimensional (2D) data of the carotid artery and the 3-dimensional (3D) simulated echocardiographic data. The qualitative and quantitative results show that images, which are reconstructed by the proposed Spatio-Temporal method have a far low error and so much better quality than those that reconstructed by conventional spatial compressive sensing method. The proposed approach via the learned overcomplete dictionary in temporal and spatial direction increases the frame rate based on the different subsampling rates. For instance, the frame rate up to two times the original sequence is achievable, while Root Mean Square Error (RMSE) is approximately 1.5 and 3 for 2D and 3D data, respectively.