Data sampling and constrained reconstruction for high-dimensional MRI
Abstract/Contents
- Abstract
- Magnetic resonance imaging (MRI) is a powerful and flexible modality for medical imaging. Data samples in MRI are acquired in the spatial frequency domain (k-space) through a time-consuming and inherently sequential process, and consequently, the scan time is proportional to the number of acquired samples. Many clinical MRI applications require data acquisition with additional dimensions, such as channels in a phased-array receive coil in parallel MRI, time in dynamic contrast-enhanced (DCE) MRI, or multiple slices excited in imaging near metallic implants, which introduce additional acquisition requirements. Modern higher-dimensional MRI applications have potential to overcome scan time limitations with constrained image reconstruction methods that exploit dependencies between auxiliary dimensions. Two major issues are that the required inversions can be ill-conditioned depending on k-space sampling patterns and may require explicit knowledge of inter-dimensional dependencies that is hard to obtain in practice. Thus, many applications compromise robustness to motion, temporal resolution, or spatial resolution. This thesis introduces new computational methods for k-space sampling and constrained reconstruction that push the current limits of high-dimensional MRI. To address the sampling problem, a general relationship between k-space sampling and problem conditioning is described using differential domain analysis, a tool that has roots in computer graphics. The theory is then used for on-the-fly design of adaptive k-space sampling patterns. Informed by new insights into k-space sampling, new methods for constrained reconstruction in DCE-MRI and MRI near metallic implants are introduced and evaluated in clinical imaging. A new k-space sampling trajectory is introduced in a clinical DCE-MRI protocol. This approach enables both conventional unconstrained reconstruction at lower temporal resolution or compressed sensing reconstruction at a retrospectively adjustable higher temporal resolution, providing robustness to breath hold loss or rapid contrast enhancement. Initial assessment of the technique, used in imaging over 3000 patients at Stanford, demonstrates its benefit. A similar approach is proposed to enable rapid imaging of patients with metallic implants by exploiting dependencies between excited slices. The reconstruction of on-resonance and off-resonance signals is formulated as robust principal component analysis, which exploits new sources of redundancy not describable in existing compressed sensing methods. Clinical evaluation in a small pilot study shows the feasibility of imaging near metal three-fold faster than the standard technique.
Description
| Type of resource | text |
|---|---|
| Form | electronic; electronic resource; remote |
| Extent | 1 online resource. |
| Publication date | 2017 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Associated with | Levine, Evan |
|---|---|
| Associated with | Stanford University, Department of Electrical Engineering. |
| Primary advisor | Hargreaves, Brian Andrew |
| Thesis advisor | Hargreaves, Brian Andrew |
| Thesis advisor | Pauly, John (John M.) |
| Thesis advisor | Vasanawala, Shreyas |
| Advisor | Pauly, John (John M.) |
| Advisor | Vasanawala, Shreyas |
Subjects
| Genre | Theses |
|---|
Bibliographic information
| Statement of responsibility | Evan Levine. |
|---|---|
| Note | Submitted to the Department of Electrical Engineering. |
| Thesis | Thesis (Ph.D.)--Stanford University, 2017. |
| Location | electronic resource |
Access conditions
- Copyright
- © 2017 by Evan Gregory Levine
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...