Accelerated Cardiovascular Magnetic Resonance Imaging Using Radial Acquisition with Compressed Sensing

Abstract

Magnetic resonance imaging (MRI) is a medical imaging technique that can visualize internal organs for examination and diagnosis. It is non-invasive and ionizing radiation-free, and provides good contrast between different soft tissues. One of the drawbacks of MRI is its lengthy acquisition time. Significant research has been done in order to accelerate the MRI data acquisition and reduce the scan time. Compressed sensing (CS) has been recently proposed for accelerating MRI acquisition time. Compressed sensing recovers the desired image from undersampled MRI dataset by exploiting the sparsity of MR image in transform domain. In this thesis, we propose CS reconstruction methods in various cardiovascular MRI applications for accelerated imaging. We consider 3D whole-heart coronary MRI. Isotropic 3D radial trajectories allow undersampling of k-space in all three dimensions, enabling accelerated acquisition of volumetric data. Our CS based approach provides further acceleration by removing undersampling artifacts and improving image quality. However, the underlying heavy computational overhead of this method is also a limiting factor which depreciates the applicability of CS. A parallelized implementation of an iterative CS reconstruction for 3D radial acquisitions using a graphics processing unit is presented to reduce the reconstruction time. The efficacy of CS is also investigated in cardiac cine MRI. Cardiac function is usually assessed using segmented cine acquisition with multiple breath-holds (BHs). Subjects are given resting periods between adjacent BHs, where no data is acquired, resulting in low acquisition efficiency. We propose an accelerated radial acquisition for BH cine imaging which utilizes the resting period to acquire additional free-breathing (FB) data without increasing scan time. The difference image between BH and FB acquisitions is used as the sparsity regularization of the CS reconstruction. Compressed sensing can be used as a respiratory motion correction technique in FB whole-heart MRI. Respiratory motion causes aliasing artifacts and blurring on the resulting image. To obtain motion-free images, a respiratory navigator is often used to track the heart position, but the scan efficiency is reduced to 30-50% resulting in a prolonged scan. We propose a CS reconstruction to correct respiratory motion by using an undersampled dataset which only contains motion-free k-space lines whereas the motion-corrupted lines are excluded from the reconstruction.