SLIM: spectral localization by imaging.

Nonspectroscopic magnetic resonance (MR) imaging often shows that a slice is composed of several compartments, each of which can be assumed to have a spatially homogeneous magnetic resonance spectrum, e.g., a limb composed of fat, muscle, bone marrow, and tumor. We show how to use structural information from such a nonspectroscopic image in order to increase the efficiency of subsequent localized spectroscopic measurements. Specifically, knowledge of the boundaries of N compartments makes it possible to reconstruct compartmental spectra from spectroscopic signals from an entire cross section with N or more different degrees of phase encoding. Experimental studies of a two-compartment phantom show that this method (SLIM) can be used to derive regional hydrogen spectra of a single slice from signals with as few as 2 phase-encoding steps, although Fourier transform chemical-shift imaging requires 64 steps to achieve a result of comparable accuracy. SLIM required only 16 phase-encoding steps to obtain accurate regional single slice spectra in a human limb with three compartments. Spectra of similar quality, obtained by Fourier transform chemical-shift imaging, required 256 to 1024 steps.