Quantifying 1H decoupled in vivo 31P brain spectra

Our objective was to develop a precise method for quantification of in vivo proton decoupled 31P spectra from the human brain. This objective required that an appropriate spectral model be created and that the quantification was performed using a non‐subjective fitting technique. The precision of the quantification was assessed using Cramér–Rao standard deviations and compared using two different spectral models: one containing a pair of peaks representing 2,3‐diphosphoglycerate, the other excluding this metabolite. The data was quantified using a Marquardt–Levenberg (ML) algorithm incorporating prior knowledge with a Hankel singular value decomposition (HSVD) performed initially to provide parameter estimates for the ML algorithm. Quantification was performed on two different in vivo 2‐D CSI 31P data sets: the first examined 11 normal controls, the second examined a single individual six times. Spectra from a region in the parieto‐occipital cortex were analyzed. The Cramér–Rao standard deviations were significantly lower for some metabolites with 2,3‐diphosphoglycerate in the model: in the repeat study mobile phospholipids (p = 0.045) and phosphocholine (p = 0.034), and in the 11 controls mobile phospholipids (p = 0.003) and Pi (p = 0.002). Copyright © 1999 John Wiley & Sons, Ltd.