Abstract
Magnetic resonance imaging (MRI) provides a safe and noninvasive method for detecting and diagnosing coronary artery disease (CAD). It can detect atherosclerotic plaque by visualizing the coronary artery wall. Changes in artery wall thickness may indicate the presence of CAD. Two-dimensional (2D) single-slice magnetic resonance (MR) coronary artery wall imaging studies have shown the thickening and occlusion of coronary vessel walls using 1.5T scanners . Three-dimensional (3D) MR coronary artery wall imaging is a very challenging problem, because the resolution must be high enough to visualize the vessel wall and the blood signal must be suppressed in the entire imaging volume. The acquisition time is long for 3D, and artifacts caused by respiratory motion during the scan also need to be minimized.
This dissertation describes research focused on the following three goals related to 3D MR coronary artery wall imaging: 1) To test and develop parallel imaging techniques that can be used to shorten the acquisition time to a breath-hold to avoid respiratory motion artifacts; 2) To develop a novel black blood technique and compare its performance to traditional methods; and 3) To develop a 3D pulse sequence with black blood preparation and parallel imaging and test its performance in coronary artery wall studies using a 3T MR scanner.
Parallel imaging is a powerful tool for accelerating MR image acquisition by using the spatial encoding inherent in an array of receiver coils. When using large coil arrays, traditional coil-by-coil parallel image reconstruction requires significant computation. In this dissertation research, we tested and improved the synthetic target technique, a new parallel imaging reconstruction algorithm that requires significantly less computation. It uses a single step in the image reconstruction to form an unaliased composite image, rather than generating an unaliased image corresponding to each coil. Array compression and parallel computing were incorporated to further reduce the required computation time. Both phantom and in-vivo studies using a Cartesian k-space trajectory were conducted, and the results were compared with GRAPPA, a standard parallel imaging technique. The synthetic target method was shown to be able to correct aliasing for brain and cardiac images with an undersampling rate up four times (4X) with a small decrease in image quality. The method can achieve up to a 30-fold computation time reduction for 32 channel data compared to GRAPPA.
We extended the synthetic target technique to non-Cartesian k-space trajectories. We added spatial weighting (PILS) masks to the original synthetic target method for constant spiral trajectories and tested the performance with in-vivo cardiac data. For radial trajectories, a zero phase map was added as a constraint on the training process. The performance was tested with phantom data. We also tested SPIRiT, an iterative parallel reconstruction technique on radial trajectories. The improved synthetic target technique removes artifacts for both spiral and radial trajectories better than the original method. For radial trajectories, the image quality of the synthetic target results was inferior to the image quality using SPIRiT, which demonstrated that the performance of the method is dependent upon the k-space trajectory. Thus, we chose to use the SPIRiT technique for subsequent coronary wall imaging studies that used non-Cartesian trajectories, because for that application image quality is more important than computation time.
The signal from blood is typically suppressed in MR coronary artery wall imaging, which makes it easier to visualize the wall and measure its thickness. One established method for suppressing blood signal is the double inversion recovery (DIR) technique, which relies on blood flowing out of the imaged slice during the inversion time (TI). However, the DIR technique often cannot suppress the blood signal throughout a thick slab imaged using a multi-slice or 3D readout. Moreover, at a field strength of 3T, the longitudinal relaxation time (T1) of blood is lengthened and thus the required inversion time for DIR is lengthened to over 600 ms, leaving a limited period during the heartbeat for placement of the readout module. In this work, a novel black blood technique called DANTE was implemented for coronary artery wall imaging. This technique was recently developed for ungated carotid artery wall imaging, but to our knowledge this is the first use of the DANTE technique for black blood gated cardiac imaging. This method enhances the contrast between blood and tissue in the entire volume of interest with less than 300 ms preparation time. We tested its performance with a 2D multi-slice spiral GRE sequence and compared the images to images acquired using the DIR technique. Because of the shorter preparation time, DANTE sequences have more flexibility in the placement of the readout model and DANTE images have comparable signal-to-noise ratio (SNR) to DIR images and higher contrast-to-noise ratio (CNR).
Free-breathing studies are commonly been for 3D coronary artery wall imaging. However, even with respiratory gating and motion compensation, respiratory motion can still lead to image blur in MR coronary artery wall imaging, causing overestimation of the wall thickness. Breath-held scans typically have less motion blur than free breathing scans, but require an efficient pulse sequence for 3D scanning. We implemented DANTE preparation, a 3D spiral readout and parallel image reconstruction to achieve 3D coronary artery wall imaging at 3T with high contrast within a breath-hold, thus minimizing motion blur. The sequence used a dual-density spiral k-space trajectory to collect parallel imaging training data and SPIRiT combined with off-resonance correction for parallel image reconstruction. Volunteer studies of right coronary artery wall imaging showed that our results have higher CNR than DIR images. They also have comparable SNR and CNR to 2D single-slice images, which indicates that parallel imaging does not cause image quality degradation. The breath-hold images have better contrast than navigator-gated results in normal volunteers, demonstrating that future coronary plaque imaging studies may benefit from 3D breath-held coronary wall imaging.
