Modeling and estimation of signals and artifacts in functional magnetic resonance imaging and computed tomography.

摘要

The objective of this dissertation is to enhance and improve two medical imaging modalities: functional Magnetic Resonance Imaging (fMRI) and x-ray Computed Tomography (CT). For image analysis in both modalities accurate modeling of the signal is important as it affects the analysis and image quality. Several components of fMRI have resisted precise modeling, due to the complexity of the signals, their underlying causes and their variations across subjects and brain regions. We introduce a maximum likelihood method that jointly estimates the Hemodynamic Response Function (HRF) and activation level in fMRI, with a regularization that allows our method to work with smaller regions of interest and therefore to better capture the variations of the HRF across the areas of the brain. Improvements in modeling achieved via the proposed methods have direct impact on enhancing the activation detection, which is a primary goal in the fMRI analysis. We also introduce extracting the general structure of the signals, such as activation level, baseline drift and the hemodynamic response function, from each data set itself using the Minimum Description Length (MDL) principle. The integration of the MDL principle with modeling the fMRI signal provides paradigm-shifting advances in the fMRI analysis. Results show that MDL-based methods offer benefits in comparison with other methods. In Computed Tomography the scatter radiation from the dose compensator causes artifacts in the image. This signal is hard to measure and it varies between each detector. This scattered radiation impacts high contrast edges such as bone-soft tissue and tissue-air interfaces, especially in brain imaging. We use Monte Carlo (MC) simulation for modeling this nuisance component of the CT data. We implement a variance reduction technique called forced detection (FD) to improve the computational efficiency. A correction algorithm for the scatter radiation is implemented which improves the image quality.