A new Bayesian approach to process chemical shift imaging data / Zhiqiang Liu.

Format:

Book

Manuscript

Microformat

Thesis/Dissertation

Author/Creator:

Liu, Zhiqiang.

Contributor:

University of Pennsylvania.

Language:

English

Subjects (All):

Penn dissertations--Structural biology.

Structural biology--Penn dissertations.

Penn dissertations--Molecular biology.

Molecular biology--Penn dissertations.

Structure Biology and Molecular Biophysics.

Academic Dissertations as Topic.

Medical Subjects:

Structure Biology and Molecular Biophysics.

Academic Dissertations as Topic.

Local Subjects:

Penn dissertations--Structural biology.

Structural biology--Penn dissertations.

Penn dissertations--Molecular biology.

Molecular biology--Penn dissertations.

Physical Description:

x, 119 pages : illustrations ; 29 cm

Production:

1997.

Summary:

As a useful technique in clinical NMR spectroscopic studies, Chemical Shift Imaging (CSI) suffers from limited spatial resolution because of low signal to noise ratio (SNR) in the data and the acquisition time constraint in the clinical settings. The conventional Fourier transform (FT) of data from k-space to the spatial domain has severe ringing and partial volume artifacts due to truncation and limited sampling, which makes quantification difficult. Although some filters or window function methods such as Hanning filtering are commonly used for reducing artifacts, they worsen the spatial resolution. Most non-FT CSI processing techniques have failed because they are unable to handle low SNR data or for practical and computational considerations. As a possible alternative to Fourier transformation, the Bayesian approach provides a better and more stable way of reconstructing low SNR data. It can often reduce the artifacts by incorporating a priori information or constraints available from high resolution proton images. Better founded mathematically, the new Bayesian method combining a probabilistic approach with Monte Carlo technique supersedes other traditional Bayesian methods such as quantified maximum entropy. The results of using this new method and combining improved data model, clearly demonstrated its promising potentials as an alternative CSI processing method. By combining this new method with other improved spectral processing techniques such as Principle Component Analysis (PCA), it will improve both the accuracy and the reliability of CSI quantification.

Notes:

Thesis (Ph.D. in Structural and Molecular Biology) -- University of Pennsylvania, 1997.

Includes bibliographical references and index.

Local Notes:

University Microfilms order no.: 98-00892.

OCLC:

244970577