Use this URL to cite or link to this record in EThOS:
Title: Multimodal Cardiovascular Image Analysis Using Phase Information.
Author: Zhang, Weiwei
ISNI:       0000 0001 3577 3407
Awarding Body: University of Oxford
Current Institution: University of Oxford
Date of Award: 2007
Availability of Full Text:
Full text unavailable from EThOS.
Please contact the current institution’s library for further details.
Cardiovascular heart disease (CVD) is one of the world's leading causes of death. Among the existing imaging techniques, cardiovascular magnetic imaging (CMR) and real-time three-dimensional echocardiography (RT3DUS) are receiving a lot of attention at the current time. Due to the 3D nature of the heart and its complex motion in 3D space, the RT3DUS is well-suited for 3D analysis of the cardiac disease. However, RT3DUS has lower specificity and sensitivity than the high spatial resolution CMR, which makes it difficult to interpret. This motivates research on assisting a clinician to automatically fuse the information from multiple imaging modalities for the early diagnosis and therapy of cardiac disease. This thesis establishes a framework for multimodal cardiovascular image analysis. First, we develop a (static) nonrigid registration of a RT3DUS volume slice and a CMR image. The local phase presentation of both images is utilized as an image descriptor of the 'featureness'. The local deformation of ventricles is modeled by a polyaffine transformation. The anchor points (or control points) used in the polyaffine transformation are automatically detected and refined by calculating a local misalignment measure based on phase mutual information. The registration process is built in an adaptive multi-scale framework to maximize the phase-based similarity measure by optimizing the parameters of the polyaffine transformation. Next, we explore a spatia-temporal alignment of RT3DUS and CMR sequences. The deformation field between both sequences is decoupled into spatial and temporal components. Temporal alignment is performed by re-slicing both sequences to contain the same number of frames and to make them correspond to the same temporal position using a differential registration. Spatial alignment is then carried out by extending the static nonrigid registration in a frame-by-frame manner. Landmarkbased validation shows that this new registration algorithm gives an accurate result. Finally, we proposed a registration-guided segmentation of the left ventricle in RT3DUS datasets. The image phase gradient is used as the edged indicator function. Incorporating local phase into the variational level set method without re-initialization enables a flexible initialization. This allows the co-registration of multimodal cardiovascular sequences to provide a strong prior knowledge about the shape of the left ventricle. We develop a registration-guided segmentation algorithm that efficiently converges to the object boundary of interest.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available