We have studied a series of normal subjects and patients with brain tumors, by using 31p three-dimensional chemical shift imaging to obtain localized 31p spectra of the brain. A significant proportion of brain cytosolic ATP in normal brain is not complexed to Mg , as indicated by the chemical shift 6 of the f-P resonance of ATP. The ATP (3-P METHODS Imaging/31p Chemical-Shift Imaging (CSI) Spectroscopy. Combined proton imaging/31P spectroscopic examinations of patients and healthy volunteers were done on a standard clinical imager [1.5 T Siemens (Iselin, NJ) Magnetom]. Positioning and proton imaging required 20 min; the 8 x 8 x 8 3D-CSI spectroscopic data required 17 min per twoacquisition block. For patients A, B, and C, 31P spectra were acquired by using 5-or 10-cm-diameter 31p surface coils and a 3D-CSI sequence with a sin/cos adiabatic half-passage pulse (17). For volunteers and patients D and E, spectra were acquired with a quadrature low-pass birdcage resonator and 3D-CSI with a 250-,us rectangular pulse with a low-tip angle as described (16); no spatial filtering was done. Retrospective alignment of the CSI voxels (16,18) was used to position the voxels symmetrically with respect to the falx cerebri and to place the central axial slice below the corpus callosum for normal subjects. For patients B through E, a CSI voxel was centered on the tumor. For patient A, whose tumor contained large cystic regions, the CSI voxel was centered on the viable portion of the tumor, as assessed from Gd contrast-enhanced proton images. After shifting, the data were Fouriertransformed in time and k-space domains, and all spectra were phased with the same corrections (obtained from the nonlocalized spectrum) to produce a three-dimensional array of 512 spectra. The 3D-CSI spectra were corrected for the missing initial time points of the free induction decay (FID) (16,19) to produce baseline-corrected spectra that are visually easier to inspect.Patients. Three male and two female patients with brain tumors, aged 36-70 years, were examined. All examination procedures were approved by the Investigational Review Board. The histories of the patients were as follows: case A, recurrent oligodendroglioma in the left temporal lobe, treated