Young, ACowan, BNash, MFrangi, AHunter, PMedrano Gracia, P2013-05-162013http://hdl.handle.net/2292/20497Since the invention of magnetic resonance imaging, the development of specific cardiac sequences has allowed radiologists to learn the anatomy and function of the heart in a new noninvasive fashion. The work presented in this thesis centres on the problem of comparing the mathematical models derived from these images. This is often termed building an atlas of the heart which, in this context, represents a collection of maps or compendiumof knowledge derived from a population of models. Statistical atlases are important because they define standard distributions of normality and pathology. The aim of this thesis is to investigate the power of finite-element models derived from cardiac magnetic resonance images to encode and describe such distributions. To that end, different state-of-the-art statistical analyses were applied to symptomatic and asymptomatic cohorts. Asymptomatic patients were drawn from the Multi-Ethnic Study of Atherosclerosis (MESA) and patients with myocardial infarct were provided by the Defibrillators To Reduce Risk By Magnetic Resonance Imaging Evaluation (DETERMINE) clinical trial, both available through the Cardiac Atlas Project. The main challenges of this thesis were to build standardised static and dynamic atlases of such cohorts (Chapter 3), identify and remove of bias sources (Chapter 4), extend the analysis to contour-derived models thus creating the largest atlas to date (Chapter 5), and predict and classify abnormalities from regional (Chapter 6) and global perspectives (Chapter 7). Results show that statistical finite-element modelling of the left ventricle provides a wealth of information which can significantly aid clinicians by offering an unprecedented comparison of features between populations. The clinical significance of these atlases will improve via addition of other cardiac conditions, while their accuracy will improve with cross-validation using independent and increasingly large samples.Items in ResearchSpace are protected by copyright, with all rights reserved, unless otherwise indicated. Previously published items are made available in accordance with the copyright policy of the publisher.https://researchspace.auckland.ac.nz/docs/uoa-docs/rights.htmAuckland Bioengineering instituteThesisCopyright: The Authorhttp://purl.org/eprint/accessRights/OpenAccessQ112903728