Analysis of first pass myocardial perfusion imaging with magnetic resonance

Abstract

Early diagnosis and localisation of myocardial perfusion defects is an important step in the treatment of coronary artery disease. Thus far, coronary angiography is the conventional standard investigation for patients with known or suspected coronary artery disease and it provides information about the presence and location of coronary stenoses. In recent years, the development of myocardial perfusion CMR has extended the role of MR in the evaluation of ischaemic heart disease beyond the situations where there have already been gross myocardial changes such as acute infarction or scarring. The ability to non-invasively evaluate cardiac perfusion abnormalities before pathologic effects occur, or as follow-up to therapy, is important to the management of patients with coronary artery disease. Whilst limited multi-slice 2D CMR perfusion studies are gaining increased clinical usage for quantifying gross ischaemic burden, research is now directed towards complete 3D coverage of the myocardium for accurate localisation of the extent of possible defects. In 3D myocardial perfusion imaging, a complete volumetric data set has to be acquired for each cardiac cycle in order to study the first pass of the contrast bolus. This normally requires a relatively large acquisition window within each cardiac cycle to ensure a comprehensive coverage of the myocardium and reasonably high resolution of the images. With multi-slice imaging, long axis cardiac motion during this large acquisition window can cause the myocardium imaged in different cross- sections to be mis-registered, i.e., some part of the myocardium may be imaged more than twice whereas other parts may be missed out completely. This type of mis-registration is difficult to correct for by using post-processing techniques. The purpose of this thesis is to investigate techniques for tracking through plane motion during 3D myocardial perfusion imaging, and a novel technique for extracting intrinsic relationships between 3D cardiac deformation due to respiration and multiple ID real-time measurable surface intensity traces is developed. Despite the fact that these surface intensity traces can be strongly coupled with each other but poorly correlated with respiratory induced cardiac deformation, we demonstrate how they can be used to accurately predict cardiac motion through the extraction of latent variables of both the input and output of the model. The proposed method allows cross-modality reconstruction of patient specific models for dense motion field prediction, which after initial modelling can be use in real-time prospective motion tracking or correction. In CMR, new imaging sequences have significantly reduced the acquisition window whilst maintaining the desired spatial resolution. Further improvements in perfusion imaging will require the application of parallel imaging techniques or making full use of the information content of the ¿-space data. With this thesis, we have proposed RR-UNFOLD and RR-RIGR for significantly reducing the amount of data that is required to reconstruct the perfusion image series. The methods use prospective diaphragmatic navigator echoes to ensure UNFOLD and RIGR are carried out on a series of images that are spatially registered. An adaptive real-time re-binning algorithm is developed for the creation of static image sub-series related to different levels of respiratory motion. Issues concerning temporal smoothing of tracer kinetic signals and residual motion artefact are discussed, and we have provided a critical comparison of the relative merit and potential pitfalls of the two techniques. In addition to the technical and theoretical descriptions of the new methods developed, we have also provided in this thesis a detailed literature review of the current state-of-the-art in myocardial perfusion imaging and some of the key technical challenges involved. Issues concerning the basic background of myocardial ischaemia and its functional significance are discussed. Practical solutions to motion tracking during imaging, predictive motion modelling, tracer kinetic modelling, RR-UNFOLD and RR-RIGR are discussed, all with validation using patient and normal subject data to demonstrate both the strength and potential clinical value of the proposed techniques.