This item is under embargo and not available online per the author's request. For access information, please visit http://libanswers.wustl.edu/faq/5640.
The qBOLD MR Model: Uncertainty in Parameter Estimation, Inclusion of Multiple Compartments and Correction of Macroscopic-Field-Inhomogeneity Effects
Date of Award
Doctor of Philosophy (PhD)
The quantitative blood oxygenation level dependent (qBOLD) imaging technique provides an MRI-based method to measure hemodynamic parameters such as oxygen extraction fraction (OEF) and deoxyhemoglobin-containing (veins and ravenous section of capillaries) cerebral blood volume fraction (dCBV). qBOLD is based on a theory of MR signal dephasing in the presence of a blood vessel network and is carried out in concert with a specialized pulse sequence -- Gradient Echo Sampling of Spin Echo (GESSE). It has been validated on phantoms and animals. In vivo human studies have also demonstrated the feasibility of this approach but also revealed that obtaining reliable results requires high SNR in the data.
Employing a Bayesian probability theory framework, we have analyzed in detail the uncertainties associated with qBOLD parameter estimates. Specifically, we have examined how the estimated OEF and dCBV depend on their "true values", signal-to-noise ratio, and data sampling strategy. Based on this analysis, we developed strategies for optimization of the qBOLD technique for dCBV and QEF evaluation. We have tested our theory on a phantom mimicking the structure of blood vessel network. (A 3D GESSE pulse sequence was used for data acquisition with subsequent analysis by Bayesian methods. The experimental results demonstrated good agreement with theoretical predictions.
The initial formulation of the qBOLD theory (He and Yablonskiy) recognized two extravascular compartments (intra- and extra-cellular) in the MR signal model. However, numerous publications treat the BOLD signal within the framework of a single compartment model, which can introduce bias if quantitative results are desired. We have obtained direct experimental evidence that the GESSE MRI signal cannot be modeled by single tissue component. Rather, a signal model incorporating a distribution of transverse relaxation rate constants and chemical shifts should be employed. We introduced a phenomenological means for describing the MR signal due to such distributions, resulting in a reduced bias in OEF measurement. In vivo studies demonstrated good qualitative agreement when compared with PET-based OEF determinations.
The MRI signal is also modulated by macroscopic magnetic field inhomogeneities. OEF values estimated by qBOLD are sensitive to such inhomogeneities. We have employed a point spread function to approximate the condition of truncated k-space and field inhomogeneity. This strategy is based on redefinition of the k-space data, which involves both a static gradient due to susceptibility effects and the imaging gradients. It was found that images reconstructed via inverse Fourier transform contain artifacts due to signal leakage, which are shift variant. Numerical simulations and phantom studies confirmed the effect of MRI signal modulation and leakage. Based on this theory, natural T2* images can be calculated for phantom and human brain.
Chair and Committee
Mark S Conradi
Joseph J Ackerman, Mark G Alford, Mark S Conradi, Joel R Garbow, James G Miller, Jason C Woods
Wang, Xiaoqi, "The qBOLD MR Model: Uncertainty in Parameter Estimation, Inclusion of Multiple Compartments and Correction of Macroscopic-Field-Inhomogeneity Effects" (2012). Arts & Sciences Electronic Theses and Dissertations. 287.