The combination of volumetric MRI and multivoxel localized MR spectroscopic imaging (MRSI) data provides the potential for quantifying variations in tissue morphology and function. These techniques have numerous applications for the evaluation of neurological diseases. While state-of-the-art whole-body MR scanners are able to acquire both types of data, there have been relatively few reports which have presented clinical applications of the technology. One of the reasons for this has been the need to develop software for reconstruction and reliable interpretation of 3D imaging and spectral data. In this article, we describe the strategy developed for quantitative analysis of combined MRI and MRSI examinations from patients with brain tumors and evaluate the key components of this procedure using both simulations and empirical data from phantoms, normal volunteers, and patients. Important factors are the use of volume or interleaved multislice anatomic images as a reference, the reconstruction and correction of the spectral data for frequency, phase, and baseline distortions and consideration of the characteristics of the coil, RF pulses used for spatial selection, and phase encoding procedures. These studies show that the metabolic parameters most critical for distinguishing tumor from normal and necrotic tissue were relative levels of choline and NAA. Levels of creatine and lactate were found to be variable, both in terms of their spatial distribution within individual lesions and between different patients.