Cardiac reconstruction and analysis from high resolution CT images

Heart disease is a major cause of mortality worldwide. Detecting/diagnosing such diseases in their early stages is critical, and heavily depends on non-invasive imaging methods, e.g., computed tomography (CT) and magnetic resonance imaging (MRI). High resolution cardiac CT imaging technology has the resolution to reveal the complex endocardial structures of the left ventricle, such as the trabeculae and the papillary muscles. However, the development of suitable methods for the quantitative analysis of these dense data sources has lagged greatly behind the development of the imaging methods themselves. As a result, in clinical practice, and in much of the research that uses these imaging data, the quantitative analysis of cardiac function has largely been confined to the calculation of simple measures of global function, such as the ejection fraction, while local function being just qualitatively assessed. Therefore there is a large amount of the functional information potentially available from cardiac images essentially untouched. In this thesis, for the first time, we extract clinical meaningful endocardial information of the left ventricle (LV) from high resolution CT images. The reconstructed result captured the fine detailed structures, which are extremely challenging to segment due to their delicate and complex nature in both geometry and topology. Our algorithm is especially designed to segment those structures, by calculating the potential missing topological structures. Using techniques from computational topology, e.g. persistent homology, our algorithm finds topological handles which are likely to be the true signal. Each handle is evaluated independently based in its saliency, rather than absolute intensities. The final segmentation with handles restored, leads to high quality segmentation of the complex structures. The initialized model was then deformed to other frames to reconstruct the 4D motion. Based on the reconstructed results, we study both the morphology and motion of the trabeculae and the papillary muscles. Proposed measurements have been clinically evaluated.Ph.D.Includes bibliographical referencesby Mingchen Ga