A subspace approach to high-resolution magnetic resonance spectroscopic imaging

With its unique capability to obtain spatially resolved biochemical profiles from the human body noninvasively, magnetic resonance spectroscopic imaging (MRSI) has been recognized as a powerful tool for in vivo metabolic studies. However, research and clinical applications of in vivo MRSI have been progressing more slowly than expected. The main reasons for this situation are the problems of long data acquisition time, poor spatial resolution and low signal-to-noise ratio (SNR) for this imaging modality. In the last four decades, significant efforts have been made to improve MRSI, resulting in a large number of fast pulse sequences and advanced image reconstruction methods. However, the existing techniques have yet to offer the levels of improvement in imaging time, spatial resolution and SNR necessary to significantly impact in vivo applications of MRSI. This thesis work develops a new subspace imaging approach to address these technical challenges to enable fast, high-resolution MRSI with high SNR. The proposed approach, coined SPICE (Spectroscopic Imaging by Exploiting Spatiospectral Correlation), is characterized by using a subspace model for integrative data acquisition, processing and image reconstruction. More specifically, SPICE represents the spectroscopic signals in MRSI using the partial separability (PS) model. The PS model implies that the high-dimensional spectroscopic signals reside in a low-dimensional subspace, which enables the design of special sparse sampling strategies for accelerated spatiospectral encoding and special image reconstruction strategies for determining the subspace and reconstructing the underlying spatiospectral function of interest from the sparse data. Using the SPICE framework, new data acquisition and image reconstruction methods are developed to enable high-resolution 1H-MRSI of the brain. We have evaluated SPICE using theoretical analysis, numerical simulations, phantom and in vivo experimental studies. Results obtained from these experiments demonstrate the unprecedented capability of SPICE in achieving accelerated MRSI with simultaneously very high resolution and SNR. We expect SPICE to provide a powerful tool for in vivo metabolic studies with many exciting applications. Furthermore, the SPICE framework also presents new opportunities for future developments in subspace-driven signal generation, signal encoding, data processing and image reconstruction methods to advance the research and clinical applications of high-resolution in vivo MRSI.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2015

Author dc:creator

• Lam, Fan

Contributors dc:contributor

• Liang, Zhi-Pei
• Boppart, Stephen A.
• Do, Minh N.
• Sutton, Brad P.

• Magnetic resonance spectroscopic imaging
• Partial separability
• Subspace modeling
• Low-rank model
• Sparse sampling

Statement dc:rights

• Copyright 2015 Fan Lam

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/78611

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/78611