A Free-Breathing Approach to High-Resolution Magnetic Resonance Imaging of Myocardial Perfusion: Fully Automated Motion Compensation with Navigators and Highly Accelerated Imaging Using k-t PCA

Cardiovascular magnetic resonance (CMR) imaging has emerged as an important research tool for studies of ischemic heart disease due to its versatility, relatively low cost, and high image quality that allows accurate and reproducible quantification of cardiac structure and function. However, while imaging of cardiac function and myocardial viability (and to some extend imaging of the coronary arteries) has reached the level of clinical acceptance, reliable and high-resolution measurements of myocardial perfusion remain difficult to obtain in practise. Further development of this technique is necessary in order for CMR to gain wide acceptance as a clinical tool for the diagnosis and management of ischemic heart disease. The primary bottleneck of myocardial perfusion imaging is the respiratory movement of the heart which hampers perfusion measurements and prevents the use of modern imaging speed-up techniques, such as the k-t BLAST method. The accuracy of perfusion measurements is significantly improved by manual motion correction, but for clinical applications the correction must be fully automated. Also the respiratory induced image artifacts seen in k-t BLAST are not readily correctable in the present implementation of the method. This thesis presents a fully automated strategy for correcting respiratory motion in myocardial perfusion CMR, and the underlying theory of k-t BLAST is revised to reduce its susceptibility to respiratory motion artifacts. The proposed respiratory motion correction is divided into two separate steps, one of which operates in real-time to deal mainly with through-plane motion and one that corrects residual in-plane motion retrospectively. Both methods are based on predictive motion modelling and use respiratory navigators as input. The modification of k-t BLAST does not comprise any explicit motion compensation. Rather, the problematic high temporal frequencies of respiratory motion are more accurately modelled by constraining the reconstruction using principal component analysis (PCA). The method is named k-t PCA. With the use of navigators, respiratory motion is compensated along all three spatial dimensions, and consequently the accuracy of myocardial perfusion quantification is improved. More importantly, all necessary steps to perform the motion correction become an integral part of the data acquisition and image reconstruction, thus promoting the clinical feasibility of myocardial perfusion CMR in general. In addition, the presented k-t PCA technique is less susceptible to respiratory motion artifacts than k-t BLAST, thereby permitting higher reduction in data acquisition. Altogether, the methods presented in this thesis provide a single-push-button CMR approach to high-resolution imaging of myocardial perfusion. This may advance CMR as a widespread clinical tool for the evaluation of ischemic heart disease.