Compressed sensing based super-resolution ultrasound imaging (SRUI)
Unbreakable ground truths govern imaging modalities: 1) imaging depth and resolution are inversely proportional; and 2) resolution of an imaging system is bounded to about half the wavelength of sound waves or light, defined as the diffraction-limit. Revolutionizing a traditional perception of low spatial resolution and deep penetration intrinsic to ultrasound modality, modern ultrasound contrast agents (e.g. microbubbles [MB]), enable an unprecedented super-resolution ultrasound miaging (SRUI). The resolution of SRUI is directly related to the size of MB (less than 10 µm). Compared to the resolution of 8 MHz ultrasound waves with the penetration depth of 100 mm, SRUI provides great improvement; 190 µm vs. 10 µm. Image reconstruction of SRUI uses subtraction between two consecutive frames of densely populated fast moving MBs. Subtraction between two frames identifies the MB location while the stationary tissue and background signals cancel out. The Cramer-Rao lower bound determines the resolution limit of SRUI. For out lab's system with an L11-5v array transducer at 20 mm depth, the minimum lateral resolution is 19 µm. Considering the lateral resolution of 8 MHz ultrasound is 135 µm, lateral resolution of SRUI improved seven times. This SRUI has been a partial success to achieve SRUI with deep tissue capability.
To precisely localize moving targets, we have developed a compressed sensing (CS)-based localization algorithm. CS algorithm is applied at localization step (γ in Fig. 1). Compared to traditional 2D cross-correlation (2DCC), localization efficiency is significantly improved (top row in Fig. 1D) and a vessel-mimicking phantom was clearly visualized by CS. Singular value decomposition (SVD) removes background noise before localization, which is suitable for ultrafast plane wave imaging. AnaGVs and MegaGVs, moving in vessel-mimicking phantom, were imaged with improved contrast after SVD.
Further reading:
To precisely localize moving targets, we have developed a compressed sensing (CS)-based localization algorithm. CS algorithm is applied at localization step (γ in Fig. 1). Compared to traditional 2D cross-correlation (2DCC), localization efficiency is significantly improved (top row in Fig. 1D) and a vessel-mimicking phantom was clearly visualized by CS. Singular value decomposition (SVD) removes background noise before localization, which is suitable for ultrafast plane wave imaging. AnaGVs and MegaGVs, moving in vessel-mimicking phantom, were imaged with improved contrast after SVD.
Further reading:
- Kim, J., Wang, Q., Zhang, S., Yoon, S., (2021). Compressed sensing-based super-resolution ultrasound imaging for faster acquisition and high quality images. IEEE Trans Biomedical Engineering. Online ahead of print. doi: https://doi.org/10.1109/TBME.2021.3070487
Figure 1. (A) Vantage 256 imaging system (B) Overall process for SRUI imaging reconstruction (C) and (D) shows improvements of image quality and point target localization by CS algorithm
Figure 2. Visualization of microvasculature of mouse tumor model by CS-based SRUI.