Faster MRI with Sparse Sampling and Low-Rank Modeling Current clinical and research MRI applications are frequently limited by the slow speed and high cost of MRI data acquisition. Faster MRI technology would enable clinicians and researchers to acquire image datasets that are substantially higher quality and more comprehensive than is currently possible with modern imaging protocols. The guiding hypothesis of this project is that substantial improvements in MRI speed can be achieved by combining sparse sampling methods with novel structured low-rank subspace modeling approaches. We have recently developed a novel low-rank subspace modeling framework (low-rank modeling of local k-space neighborhoods, or ?LORAKS?), which has the unique feature that it can be used to model multiple standard MRI constraints (sparse image support, smooth phase, and parallel imaging) in a calibration-free way. We have also demonstrated that LORAKS can enable high-quality MRI images to be reconstructed from much less data than required by the conventional sampling theorem. In contrast to many of the alternative reconstruction methods for undersampled data, LORAKS offers high acceleration rates and unprecedented compatibility with a wide variety of MRI data sampling schemes. The objective of this project is the further investigation and development of LORAKS for practical MRI neuroimaging applications.
Aim 1 involves the practical implementation and optimization of LORAKS for high-resolution anatomical imaging, and includes pulse sequence development, sampling-scheme optimization, development of publicly-distributed software, and an empirical evaluation of the impact of accelerated MRI on practical human brain morphometry studies.
Aim 2 involves the extension of the LORAKS framework to advanced fast imaging protocols that utilize simultaneous multi-slice data acquisition and echo-planar imaging strategies.
This aim i ncludes further mathematical generalization of the LORAKS framework, pulse sequence development, exploration and optimization of novel sampling schemes, development of publicly-distributed software, and evaluations in the context of quantitative diffusion MRI. The methods developed through this work are expected to substantially enhance the performance of accelerated MRI, and are general enough that they can be used to enhance virtually any clinical MRI application. This work will also lay important groundwork for future studies of accelerated MRI.

Public Health Relevance

Faster MRI with Sparse Sampling and Low-Rank Modeling Magnetic resonance imaging (MRI) is an important non-invasive imaging modality that can provide valuable insights into patient-specific disease processes, but is limited in clinical scenarios by the slow speed and high cost of MRI data acquisition. This project will develop novel fast MRI technology that would enable clinicians and researchers to acquire image datasets that are substantially higher quality and more comprehensive than is currently possible with modern imaging protocols.

Agency
National Institute of Health (NIH)
Institute
National Institute of Biomedical Imaging and Bioengineering (NIBIB)
Type
Exploratory/Developmental Grants (R21)
Project #
1R21EB022951-01
Application #
9168254
Study Section
Biomedical Imaging Technology A Study Section (BMIT-A)
Program Officer
Liu, Guoying
Project Start
2016-07-01
Project End
2018-04-30
Budget Start
2016-07-01
Budget End
2017-04-30
Support Year
1
Fiscal Year
2016
Total Cost
$194,416
Indirect Cost
$69,416
Name
University of Southern California
Department
Engineering (All Types)
Type
Schools of Engineering
DUNS #
072933393
City
Los Angeles
State
CA
Country
United States
Zip Code
90032
Lobos, Rodrigo A; Kim, Tae Hyung; Hoge, W Scott et al. (2018) Navigator-Free EPI Ghost Correction With Structured Low-Rank Matrix Models: New Theory and Methods. IEEE Trans Med Imaging 37:2390-2402
Kim, Tae Hyung; Bilgic, Berkin; Polak, Daniel et al. (2018) Wave-LORAKS: Combining wave encoding with structured low-rank matrix modeling for more highly accelerated 3D imaging. Magn Reson Med :
Bilgic, Berkin; Kim, Tae Hyung; Liao, Congyu et al. (2018) Improving parallel imaging by jointly reconstructing multi-contrast data. Magn Reson Med 80:619-632
Varadarajan, Divya; Haldar, Justin P (2017) A theoretical signal processing framework for linear diffusion MRI: Implications for parameter estimation and experiment design. Neuroimage 161:206-218
Kim, Daeun; Doyle, Eamon K; Wisnowski, Jessica L et al. (2017) Diffusion-relaxation correlation spectroscopic imaging: A multidimensional approach for probing microstructure. Magn Reson Med 78:2236-2249