RUcore Resource Object
RUcore Resource Object
PDF-1(4.47 MB)
Uniform TitleModel-based image segmentation in medical applications
NameQian, Zhen (author), Metaxas, Dimitris (chair), Pavlovic, Vladimir (internal member), Elgammal, Ahmed (internal member), Rutgers University, Graduate School - New Brunswick,
Degree Date2007
Date Created2007
SubjectComputer Science,
Diagnostic imaging,
Image processing,
Magnetic resonance imaging
DescriptionImage segmentation is an essential and indispensable step in medical image analysis. It partitions the image into meaningful anatomic or pathological structures. Because medical image segmentation needs high level medical and anatomic knowledge, model-based segmentation methods are highly desirable. In this thesis, we will first give a short survey of current approaches of medical image segmentation. Then we specifically develop appearance and shape models for different segmentation tasks. These models are either obtained from visual observation and prior human expertise, or from certain automatic
machine learning methods. In this thesis, two model-based image segmentation algorithms are developed for 3D MR colonography and 2D cardiac tagged MRI. For 3D MR colonography, we manually build the shape and intensity model. For 2D tagged MRI, we learn the shape and local appearance model from a training set. In each application, besides the models, we give complete details in solving the segmentation problems, such as how we correct the MR image intensity inhomogeneity and how we automatically initialize the segmentation. Both model-based methods perform well on real medical image data.
machine learning methods. In this thesis, two model-based image segmentation algorithms are developed for 3D MR colonography and 2D cardiac tagged MRI. For 3D MR colonography, we manually build the shape and intensity model. For 2D tagged MRI, we learn the shape and local appearance model from a training set. In each application, besides the models, we give complete details in solving the segmentation problems, such as how we correct the MR image intensity inhomogeneity and how we automatically initialize the segmentation. Both model-based methods perform well on real medical image data.
NoteM.S.
NoteIncludes bibliographical references (p. 58-60).
Genretheses
Persistent URLhttp://hdl.rutgers.edu/1782.2/rucore10001600001.ETD.16760
LanguageEnglish
CollectionGraduate School - New Brunswick Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.