In vivo spectroscopic magnetic resonance imaging using estimation theory

The ability to map spectroscopic components for in vivo application is a highly desirable goal. Thusfar it has been unavailable because of the low SNR inherent in the measurement of each voxel. In this paper we deal with this low SNR in two ways. First, estimation theory, using a priori data, is used to estimate the amplitudes in each voxel. Second, the resultant estimates are presented in an image format so that they are readily correlated with anatomical and physiological patterns. A computer simulation is presented of a case where the SNR is‐10 dB. At this level, conventional Fourier transform spectroscopy provides meaningless results. Using estimation theory and an imaging format, a simulated lesion is readily seen. © 1986 Academic Press, Inc.