Journal of magnetic resonance vol:132 issue:2 pages:197-203
Quantification of metabolites in H-1 spectra is difficult because of the presence of an unwanted water signal. Preprocessing, or removing the water contribution of a H-1 spectrum, in the time domain is usually done using the state-space approach HSVD. HSVD removes the residual water and its side lobes, thereby reducing the baseline for the metabolites of interest and allowing subsequent data analysis using more sophisticated nonlinear least squares algorithms. However, the HSVD algorithm is computationally expensive because it estimates the signal subspace using the singular value decomposition (SVD). We show here that replacing the SVD by a low-rank revealing decomposition speeds up the computations without affecting the accuracy of the wanted parameter estimates. (C) 1998 Academic Press.