Information on tissue metabolites obtained by in vivo proton magnetic resonance spectroscopy (MRS) offers considerable potential for clinical diagnosis and biomedical research. Analysis of this data benefits from parametric modeling methods that use a priori spectral information. We have previously used computer simulation methods to obtain detailed spectral information, using parameters measured by high-resolution NMR of metabolites in solution. Improvements of the spectral model are needed to account for additional signal contributions, and differences in the in vivo environment or metabolite conformations. A goal of this project is accurate characterization of the NMR parameters of metabolites and other signal contributions observed with in vivo 1H MRS in brain, in order to improve the accuracy of parametric spectral analysis. Studies will be carried out using high-resolution NMR of metabolites in solution, in tissue homogenates, and in vivo. Additional studies will characterize the macromolecular contributions to spectra of brain, which will also be incorporated into the parametric signal model to further improve the spectral analysis. A second goal of this study is the development of methods for classification of large amounts of in vivo spectra, such as that obtained for clinical MR spectroscopic imaging studies using automated parametric spectral analysis. Supervised and unsupervised methods will be used to probabilistically identify each voxel with multiple possible classes. This work will extend clinical MRS applications by providing increased accuracy and information on the intracellular environment. This development will support ongoing clinical studies investigating changes of brain metabolites associated with Alzheimer's disease and aging, epilepsy, and multiple sclerosis.
Showing the most recent 10 out of 23 publications
