11 September 2015 Imaging industry expectations for compressed sensing in MRI
Author Affiliations +
Abstract
Compressed sensing requires compressible data, incoherent acquisition and a nonlinear reconstruction algorithm to force creation of a compressible image consistent with the acquired data. MRI images are compressible using various transforms (commonly total variation or wavelets). Incoherent acquisition of MRI data by appropriate selection of pseudo-random or non-Cartesian locations in k-space is straightforward. Increasingly, commercial scanners are sold with enough computing power to enable iterative reconstruction in reasonable times. Therefore integration of compressed sensing into commercial MRI products and clinical practice is beginning. MRI frequently requires the tradeoff of spatial resolution, temporal resolution and volume of spatial coverage to obtain reasonable scan times. Compressed sensing improves scan efficiency and reduces the need for this tradeoff. Benefits to the user will include shorter scans, greater patient comfort, better image quality, more contrast types per patient slot, the enabling of previously impractical applications, and higher throughput. Challenges to vendors include deciding which applications to prioritize, guaranteeing diagnostic image quality, maintaining acceptable usability and workflow, and acquisition and reconstruction algorithm details. Application choice depends on which customer needs the vendor wants to address. The changing healthcare environment is putting cost and productivity pressure on healthcare providers. The improved scan efficiency of compressed sensing can help alleviate some of this pressure. Image quality is strongly influenced by image compressibility and acceleration factor, which must be appropriately limited. Usability and workflow concerns include reconstruction time and user interface friendliness and response. Reconstruction times are limited to about one minute for acceptable workflow. The user interface should be designed to optimize workflow and minimize additional customer training. Algorithm concerns include the decision of which algorithms to implement as well as the problem of optimal setting of adjustable parameters. It will take imaging vendors several years to work through these challenges and provide solutions for a wide range of applications.
© (2015) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Kevin F. King, Kevin F. King, Adriana Kanwischer, Adriana Kanwischer, Rob Peters, Rob Peters, } "Imaging industry expectations for compressed sensing in MRI", Proc. SPIE 9597, Wavelets and Sparsity XVI, 959705 (11 September 2015); doi: 10.1117/12.2186023; https://doi.org/10.1117/12.2186023
PROCEEDINGS
6 PAGES


SHARE
RELATED CONTENT

Multi-contrast magnetic resonance image reconstruction
Proceedings of SPIE (March 19 2015)
Prior image constrained compressed sensing (PICCS)
Proceedings of SPIE (March 02 2008)
MR images from fewer data
Proceedings of SPIE (October 14 2012)
Quantifying and correcting motion artifacts in MRI
Proceedings of SPIE (September 04 2006)

Back to Top