Sequence Development and Data Processing of Echo Planar Chemical Shift Imaging

Chemical Shift Imaging (CSI) methodology has been used extensively to obtain metabolic data from a wide variety of organs in both humans and animals. Its spectroscopy and imaging components make possible the localized analysis of metabolic information. However, conventional CSI uses dedicated phase encoding (PE) gradients, which makes it slow for clinical use. Echo Planar Chemical Shift Imaging (EPCSI) acquires one spatial and the spectral data in a single readout. Therefore, it has the potential to substantially reduce scan time by eliminating one PE gradient. However, there are still several problems with EPCSI that prevent it from wide clinical utility. First, it is technically challenging to implement an EPCSI sequence with adequate spatial resolution and spectral bandwidth on certain available commercial MR platforms. Second, due to the low concentration of the metabolites, it is often essential to have adequate water and fat suppression. However, these suppression techniques are conventionally included in the spectroscopy mode, not the imaging mode on Philips platform. Therefore robust water and fat suppression need to be implemented in our EPCSI that is developed from the EPI sequence on the PHILIPS scanner. Third, EPCSI generates a huge amount of related spectra data, which makes a robust post-processing method for EPCSI imperative. In this thesis work, we aim to implement a robust EPCSI sequence on the PHILIPS platform and test its potential application by applying it to high resolution water spectral analysis and low resolution multi-spectral analysis. The sequence we developed is capable of achieving more than 128 x 128 pixel spatial resolution or up to 1650 Hz spectral bandwidth and 1024 spectral points. The typical water suppression ratio acquired with CHESS is more than 500. The implementation of an improved fat suppression provides a maximum of 8 regional suppression technique (REST) slabs. For post-processing, a comprehensive data analysis approach that includes principal component analysis (PCA) followed by constrained Nonnegative Matrix Factorization (cNMF) has been developed. We demonstrate the ability of PCA to correct for undesirable variations in the spectra caused by factors such as field inhomogeneity so that meaningful information about the heterogeneous subvoxelar magnetic microenvironments can be revealed. We also show cNMF as a valuable tool for finding constituent spectra, as the principal components (PCs) calculated from PCA do not represent the true spectral peak. To demonstrate its application, we applied our EPCSI sequence and post-processing package to several experiments. We show good spectral quality can be obtained on both phantom and healthy volunteers with our method; high resolution metabolic information can be obtained from multiple chemicals simultaneously by using EPCSI combined with PCA. We applied the high resolution water spectral analysis on both healthy young volunteers and older volunteers with white matter hyperintensity (WMH), and observed clear structure of gray matter, white matter and also WMH in the score maps of the first few PCs and the abundance maps of the first few constituent spectra which can explain some underlying physiological causes. Finally, our EPCSI sequence and post-processing also have the potential for functional MR studies and may explain the mechanism of negative BOLD signal.