Differentiation of metastases from high-grade gliomas using short echo time 1H spectroscopy

PURPOSE: To determine if short echo time (TE) (1)H magnetic resonance spectroscopy (MRS) can distinguish between intracranial metastases and glioblastomas. MATERIALS AND METHODS: TE 30-msec spectra were acquired (1.5 T) from voxels entirely within tumors from 23 glioblastoma patients and 24 metastases patients (3 breast carcinomas, 1 bladder carcinoma, 8 lung carcinomas, 3 probable lung carcinomas, 6 melanomas, 1 stomach carcinoma, and 2 undetermined). Spectra were analyzed quantitatively (LCModel) to determine metabolite, lipid, and macromolecule concentrations. All tumors were previously untreated and classified histopathologically. RESULTS: The lipid peak area (LPA) ratio (total peak area at ca. delta1.3 to that at ca. delta0.9) was 2.6 +/- 0.6 (N = 25) for glioblastomas and 3.8 +/- 1.4 (N = 34) for metastases (P < 0.0001). There were no significant differences in metabolite or lipid concentrations between the tumor groups. The LPA ratio provided 80% sensitivity and 80% specificity for discriminating metastases from glioblastomas. CONCLUSION: Lipid and macromolecule (LM) signals can dominate (1)H spectra of high-grade tumors and have characteristics that allow significant discrimination of metastases from glioblastomas. Work is now needed to determine the source and biophysical characteristics of these LM signals to further improve differentiation by optimizing the data acquisition and analysis protocol.