Most clinical MRI applications rely on contrast-weighted images, which are complex functions of intrinsic MR tissue parameters and external scan settings. These images are qualitative in nature, providing limited capability for direct inter- and intra- patient comparisons across different institutions and/or across scanners. Although the potentials of quantitative MRI, which directly maps the underlying tissue parameters, have long been recognized, achieving this goal often requires lengthy acquisition times. Magnetic resonance fingerprinting (MRF) is a very recent breakthrough in quantitative MRI, which provides a novel data acquisition and image reconstruction paradigm to enable simultaneous quantitative measurements of multiple MR tissue parameters and experiment- specific parameters (e.g., field inhomogeneity). It dramatically shortens acquisition times to ~15s per imaging slice. However, it can still result in clinically unacceptable lengthy acquisitions of up to 20 minutes for volumetric coverage of e.g., the brain. Furthermore, this new promising imaging approach still lacks important detailed quantification of the accuracy of the measured parameters and could benefit from further optimization of its acquisition and reconstruction, both of which could be achieved through development of a rigorous statistical and signal processing framework. This proposal aims to develop a novel statistical imaging framework to optimize data acquisition and image reconstruction for MRF. On the image reconstruction side, we will develop statistically-optimal approaches to improve MR tissue parameter estimation. We will characterize the estimation performance to provide error bars for estimated tissue parameters. We will extend these reconstruction approaches to simultaneous multislice (SMS) acquisitions for SNR-efficient whole-brain coverage. Further we will incorporate with advanced image prior models (e.g., low-rank and sparse model) for reduced total acquisition time and/or improved estimation performance. On the data acquisition side, we will characterize the SNR efficiency of MRF acquisitions using estimation-theoretic metrics (e.g., Cramer-Rao bounds), and optimize acquisition parameters based on these metrics to encode tissue MR parameters into the most informative measurements. We will further extend it to optimized SMS-MRF to improve the final SNR efficiency for whole-brain imaging. Together, we aim to produce an optimized MRF technology for quantitative MRI, enabling more accurate MR tissue parameter mapping at a clinically desirable resolution (e.g., 1.0 1.0 3.0 mm3) with whole-brain coverage and ~5- 6x more efficient data acquisition (i.e., total acquisition time within 3 minutes).
This project aims to develop a highly accurate and efficient magnetic resonance fingerprinting technology for quantitative MR imaging. It will enable quantitative tissue characterization, and allows for direct inter and intra patient comparisons across different institutions and/or cross scanners. This holds great potential to revolutionize the next-generation MRI systems for improved diagnostic imaging.
| Zhao, Bo; Setsompop, Kawin; Adalsteinsson, Elfar et al. (2018) Improved magnetic resonance fingerprinting reconstruction with low-rank and subspace modeling. Magn Reson Med 79:933-942 |
| Bo Zhao; Bilgic, Berkin; Adalsteinsson, Elfar et al. (2017) Simultaneous multislice magnetic resonance fingerprinting with low-rank and subspace modeling. Conf Proc IEEE Eng Med Biol Soc 2017:3264-3268 |
