Reconstruction is key to the generation of anatomic, functional and biochemical information in the field of Magnetic Resonance (MR) in medicine. Here, prior knowledge based on various conditions is utilized through reconstruction to accelerate current MR techniques and reduce artifacts.First, prior knowledge from Magnetic Resonance Imaging (MRI) is exploited to accelerate spatial localization in Magnetic Resonance Spectroscopy (MRS). The MRS information is contained in one extra chemical shift dimension, beyond the three spatial dimensions of MRI, and can provide valuable in vivo metabolic information for the study of numerous diseases. However, its research and clinical applications are often compromised by long scan times. Here, a new method of localized Spectroscopy with Linear Algebraic Modeling (SLAM) is proposed for accelerating MRS scans. The method assumes pre-conditions that the MRS scan is preceded by a scout MRI scan and that a compartment-averaged MRS measurement will suffice for the assessment of metabolic status. SLAM builds a priori MRI-based segmentation information into the standard Fourier-encoded MRS model of chemical shift imaging (CSI), to directly reconstruct compartmental spectra.Second, SLAM is extended to higher dimensions and to incorporate parallel imaging techniques that deploy pre-acquired sensitivity information based on the use of separate multiple receive-coil elements, to further accelerate scan speed. In addition, eddy current-induced phase effects are incorporated into the SLAM model, and a modified reconstruction algorithm provides improved suppression of signal leakage due to heterogeneity in the MRS signal, especially when employing sensitivity encoding.Third, prior information from MRI is also used to reduce the problem of lipid artifacts in 1H brain CSI. CSI is routinely used for human brain MRS studies, and low spatial resolution in CSI causes partial volume error and signal ;;bleed’ that is especially deleterious to voxels near the scalp. A standard solution is to apply spatial apodization, which adversely affects spatial resolution. Here, a novel automated strategy for partial volume correction that employs grid shifting (;;PANGS’) is presented, which minimizes lipid signal bleed without compromising spatial resolution. PANGS shifts the reconstruction coordinate in a designated region of image space—the scalp, identified by MRI—to match the tissue center of mass instead of the geometric center of each voxel. Last, prior knowledge of the spatially sparse nature of endoscopic MRI images acquired with tiny internal MRI antennae, and that of the null signal location of the endoscopic probe, are used to accelerate MR endoscopy and reduce motion artifacts. High-resolution endoscopic MRI is susceptible to degradation from physiological motion, which can necessitate time-consuming cardiac gating techniques. Here, we develop acceleration techniques based on the compressed sensing theory, and un-gated motion compensation strategies using projection shifting, to effectively produce faster motion-suppressed MRI endoscopy.
【 预 览 】
附件列表
Files
Size
Format
View
Advancing Magnetic Resonance Spectroscopy and Endoscopy with Prior Knowledge