Improved analysis of 1H-MR spectra in the presence of mobile lipids

Focal brain lesions can be associated with proton magnetic resonance spectra (H-1-MRS)-detectable mobile lipids, reflecting severe tissue degradation and necrosis. However, advanced fitting procedures, such as the LCModel, fail to adequately fit spectra in the presence of lipid resonances. To overcome this, different approaches to generate lipid model spectra were compared using a phantom, real in vivo data, and simulated data. Twenty-six in vivo short-echo time (TE) H-1-MRS from 21 malignant gliomas, four infections, and one ischemia were analyzed to evaluate the performance of the modified LCModel fit. Adding simulated aliphatic resonances at 1.3 and 0.9 ppm improved the overall fitting quality remarkably and allowed good separation of lactate and alanine. Also, a better differentiation of glioblastomas and anaplastic gliomas was achieved. In conclusion, we propose a simple way to efficiently include lipid resonances in the LCModel, allowing a better fit of in vivo short-TE H-1-MRS, and demonstrate the diagnostic potential of quantitative assessment of mobile lipids in brain tumors. (C) 2001 Wiley-Liss, Inc.