Noninvasive evaluation of malignancy of brain tumors with proton MR spectroscopy

PURPOSE: To test clinical proton MR spectroscopy as a noninvasive method for predicting tumor malignancy. METHODS: Water-suppressed single-voxel point resolved spectroscopy in the frontal white matter of 17 healthy volunteers and 25 patients with brain tumors yielded spectra with peaks of N-acetyl aspartate (NAA), choline-containing compounds (Cho), creatine/phosphocreatine (Cre), and lactate. These peak intensities were semiquantitated as a ratio to that of the external reference. The validity of the semiguantitation was first evaluated through phantom and volunteer experiments. RESULTS: The variation in measurements of the designated region in the volunteers was less than 10%. Normal ranges of NAA/reference, Cho/reference, and Cre/reference were 3.59 ± 0.68, 1.96 ± 0.66, and 1.53 ± 0.64 (mean ± SD), respectively. In 17 gliomas, the Cho/reference value in high-grade gliomas was significantly higher than in low-grade gliomas. Levels of NAA/reference were also significantly different in low-grade and high-grade malignancy. In eight meningiomas (four newly diagnosed and four recurrent), the level of Cho/reference was significantly higher in recurrent meningiomas than in normal white matter or in newly diagnosed meningiomas. CONCLUSIONS: Higher grades of brain tumors in this study were associated with higher Cho/reference and lower NAA/reference values. These results suggest that clinical proton MR spectroscopy may help predict tumor malignancy.