Learning-based Algorithms for Inverse Problems in MR Image Reconstruction and Quantitative Perfusion Imaging

Magnetic Resonance Imaging (MRI) is useful in medical imaging because it is free from ionizing radiation and is able to provide excellent soft tissue contrast. However, MRI suffers from drawbacks like long scanning durations. In modalities like Arterial Spin Labeling (ASL), which is used for non-invasive and quantitative perfusion imaging, low SNR and lack of precision in parameter estimates present significant problems.

In this thesis, we first investigate the reconstruction of MR images from fewer measurements using data-driven machine learning, thereby reducing the scan duration. Specifically, we combine a supervised and an unsupervised (blind) learned dictionary in a residual fashion as a spatial prior in MR image reconstruction, and then extend this framework to include deep supervised learning. The latter, called blind primed supervised learning, proposes that there exists synergy between features learned using shallower dictionary-based and deep supervised learning-based approaches. We exploit this synergy to yield reconstructions of improved quality. Secondly, we propose a framework for providing fast and precise estimates for multiple physiological parameters from ASL scans by estimation-theory-based optimization of ASL scan design, and combination with MR Fingerprinting. For this purpose, we use the Cramer-Rao Lower Bound for scan design optimization, and deep learning for regression-based estimation.

Chair: Professors Jeffrey A. Fessler & Luis Hernandez-Garcia

Required Passcode: 986914