Original Research Analysis of the Spatial Characteristics of Metabolic Abnormalities in Newly Diagnosed Glioma Patients Xiaojuan Li, MS, 1 Ying Lu, PhD, 2 Andrea Pirzkall, MD, 3 Tracy McKnight, PhD, 1 and Sarah J. Nelson, PhD 1 * Purpose: To evaluate the role of 3D MR spectroscopic im- aging (MRSI) as a tool for characterizing heterogeneity within a lesion in glioma patients. Materials and Methods: Forty-nine patients with newly diagnosed glioma were studied with 3D water-suppressed proton (1-H) MRSI. Signal intensities from choline (Cho), creatine (Cr), N-acytel aspartate (NAA), and lactate/lipid (LL) were estimated from the spectra. Regions of interest (ROIs) corresponding to the metabolic abnormalities were defined and compared with the anatomic lesions. Results: This study showed that the tumor burden mea- sured with either the volumes of the metabolic abnormali- ties or the metabolic levels in the most abnormal voxels was correlated with the degree of malignancy of the tumor. The volumes of elevated Cho and decreased NAA were useful for distinguishing low-grade from high-grade lesions. The vol- ume of abnormal LL was correlated with the existence of necrosis and with the volume of contrast-enhancing lesions in high-grade lesions. The differences in the volume of ab- normal LL were also statistically significant between pa- tients in each grade. Conclusion: These 3D-MRSI data provide important addi- tional information to conventional MRI for evaluating and characterizing gliomas. Key Words: magnetic resonance spectroscopy; glioma; spectral quantification; metabolic abnormality J. Magn. Reson. Imaging 2002;16:229 –237. © 2002 Wiley-Liss, Inc. GLIOMAS ARE THE MOST COMMON TYPE of primary brain tumor in adults. Although the prognosis for pa- tients with high-grade gliomas is relatively poor, the choice of treatment protocol can have a significant im- pact on survival. In addition to age and performance status, the key factors in determining treatment strat- egy are tumor extent, grade, and histopathology, and response to therapy. Conventional MRI is currently ap- plied to characterize the spatial extent of the morpho- logical lesions. Although the gadolinium-enhancing le- sion indicates regions where the blood brain barrier is compromised, it is not always a reliable indicator of viable tumor due to the presence of non-enhancing tumor or contrast-enhancing necrosis. The T2- weighted hyperintensity lesion may correspond to mi- croscopic tumor infiltration, but it is not able to distin- guish tumor infiltration from regions of edema and inflammation. Functional or metabolic information us- ing alternative imaging methodologies is required for the accurate detection and evaluation of a viable tumor. In vivo MR spectroscopy (MRS) techniques, including single-voxel MRS (1– 8) and more advanced MRS imag- ing (MRSI) (9 –15), are noninvasive techniques that pro- vide information about the metabolic characteristics of tissue in the brain. Other studies have applied high- field MRS to ex vivo tissue samples, or focused on the correlation between in vivo and ex vivo results (16 –18). These studies have indicated that MRS is valuable for distinguishing active tumor from normal tissue or edema, and radiation-induced necrosis from recurrent tumor. This suggests that MRS may be important in defining the spatial extent and characteristics of tu- mors. Methods for examining spatial heterogeneity are particularly important for gliomas because the lesions may be composed of regions with different tumor grades, and even regions with the same tumor grade may respond differently to therapy. Three-dimensional (3D) MRSI is essential for the evaluation of such heter- ogeneity. With the long echo times (TE = 144 msec) used in our in vivo 1 H MRSI method, five classes of molecules are generally observed in brain spectra: 1) N-acetyl aspar- tate (NAA); 2) free choline and choline-containing com- pound, including phosphocholine and glycerophospho- choline (Cho); 3) creatine and phosphocreatine (Cr); 4) lactate (Lac); and 5) mobile lipid (Lip). The typical spec- trum corresponding to tumor shows an increased Cho peak, which corresponds to increased cell density and membrane turnover in neoplastic tissue, and a de- creased or diminished NAA peak, which indicates a loss 1 Department of Radiology, Magnetic Resonance Science Center (MRSC), University of California–San Francisco, San Francisco, Cali- fornia. 2 Department of Radiology, University of California–San Francisco, San Francisco, California. 3 Department of Radiation Oncology, University of California–San Fran- cisco, San Francisco, California. Presented in part at the 9th Annual Meeting of ISMRM, Glasgow, Scot- land, 2001. Contract grant sponsor: NIH; Contract grant numbers: R01-CA79719; R01-CA59880. *Address reprint requests to: S.J.N., Dept. of Radiology, Box 1290, 1 Irving St., A-C109, San Francisco, CA 94143. Email: nelson@mrsc.ucsf.edu Received January 15, 2002; Accepted April 24, 2002. DOI 10.1002/jmri.10147 Published online in Wiley InterScience (www.interscience.wiley.com). JOURNAL OF MAGNETIC RESONANCE IMAGING 16:229 –237 (2002) © 2002 Wiley-Liss, Inc. 229