| Summary: | <p>In response to myocardial infarction injury, the heart undergoes a remodelling process that alters chamber structure and affects cardiac function. Due to its excellent anatomical and functional tissue characterisation, cardiovascular magnetic resonance (CMR) imaging has become a well-established diagnostic tool in such cases of ischemic heart disease; in particular, CMR imaging is increasingly used to assess cardiac state in the acute phase of the disease. The longitudinal analysis of these CMR acquisitions can help to elucidate the remodelling process and evaluate the efficacy of clinical interventions.</p> <p>A limiting factor for all post-processing operations is the constrained resolution of clinically-acquired CMR data. The standard planar images obtained are highly anisotropic, leading to poor quality volume reconstructions when image data are integrated. This can have detrimental effects on operations such as segmentation and registration, reducing confidence in the clinical conclusions resulting from image analysis.</p> <p>In this thesis, we look to develop registration methods that enable the comparative analysis of longitudinally acquired CMR data, in the form of complementary cine and late gadolinium-enhanced (LGE) studies. Phase-based registration approaches are proposed to address the challenges of data misalignment observed in individual imaging sessions, as well as to compensate for the differences between longitudinal follow-up data.</p> <p>Poor quality volumetric reconstructions can have a detrimental effect on the 3D registration accuracy. Therefore, in the final part of this thesis, we present an anisotropic reaction-diffusion-based reconstruction process, which reduces some of the undesirable structure-blurring outcomes from linear interpolation. In addition, we explore a number of adaptations to conventional interpolation of short-axis slice data, in which we aim to maximise the utilisation of structural information contained in the imaging studies acquired.</p>
|
|---|