A Generalized Series Approach to MR Spectroscopic Imaging

The problem of precise spatial localization of spectral information in magnetic resonance (MR) spectroscopic imaging is addressed. A novel method, called GSLIM (generalized spectral location by imaging), is proposed to make possible the marriage of high-resolution proton imaging with spectroscopic imaging and localization. This method improves on the conventional Fourier series inversion method used in chemical shift imaging (CSI) and the compartmental modeling method used in SLIM by using a generalized series framework for optimal representation of the spectral function. In this way, a priori information extracted from proton imaging can be used, as in SLIM, and the robustness and data consistency of CSI are also retained. Simulation results show that GSLIM can significantly reduce spectral leakage in CSI and inhomogeneity errors in SLIM. It can also reveal compartmental inhomogneities, and can easily be extended to handle other a priori constraints when necessary. This approach, with some further development, may achieve an optimal combination of sensitivity, quantitative accuracy, speed, and flexibility for in vivo spectroscopy. The generalized series mathematical framework developed should also prove useful for solving other inverse problems in physics and engineering.