AJR:185, December 2005 1471 AJR 2005; 185:1471–1476 0361–803X/05/1856–1471 © American Roentgen Ray Society Weybright et al. MR Spectroscopy of Brain Tumor Versus Radiation Injury Neuroradiology • Original Research Differentiation Between Brain Tumor Recurrence and Radiation Injury Using MR Spectroscopy Patrick Weybright 1 Pia C. Sundgren 1 Pavel Maly 1 Diana Gomez Hassan 1 Bin Nan 2 Suzan Rohrer 1 Larry Junck 3 Weybright P, Sundgren PC, Maly P, et al. DOI:10.2214/AJR.04.0933 Received June 14, 2004; accepted after revision December 20, 2004. 1 Department of Radiology, University of Michigan, 1500 E Medical Center Dr., Ann Arbor, MI 48109-0030. Address correspondence to P. C. Sundgren. 2 School of Public Health, University of Michigan, Ann Arbor, MI. 3 Department of Neurology, University of Michigan, Ann Arbor, MI. OBJECTIVE. The purpose of our study was to explore the feasibility and utility of 2D chemical shift imaging (CSI) MR spectroscopy in the evaluation of new areas of contrast en- hancement at the site of a previously treated brain neoplasm. MATERIALS AND METHODS. Two-dimensional CSI (point-resolved spectroscopy sequence [PRESS]; TR/TE, 1,500/144) was performed in 29 consecutive patients (4–54 years old; mean age, 34 years) who had a new contrast-enhancing lesion in the vicinity of a previously diagnosed and treated brain neoplasm. Clinical and imaging follow-up, and histopathology in 16 patients, were used as indicators of the identity of a lesion. RESULTS. Diagnostic-quality spectra were obtained in 97% of the patients. The Cho/Cr (choline/creatine) and Cho/NAA (choline/N-acetyl aspartate) ratios were significantly higher, and the NAA/Cr ratios significantly lower, in tumor than in radiation injury (all three differ- ences, p < 0.0001). The Cho/Cr and Cho/NAA ratios were significantly higher in radiation in- jury than in normal-appearing white matter (p < 0.0003 and p < 0.0001, respectively), whereas NAA/Cr ratios were not different (p = 0.075). Mean Cho/Cr ratios were 2.52 for tumor, 1.57 for radiation injury, and 1.14 for normal-appearing white matter. Mean Cho/NAA ratios were 3.48, 1.31, 0.79, and mean NAA/Cr ratios were 0.79, 1.22, and 1.38, respectively. When values greater than 1.8 for either Cho/Cr or Cho/NAA ratios were considered evidence of tumor, 27 of 28 patients could be correctly classified. CONCLUSION. Two-dimensional CSI MR spectroscopy can differentiate tumor from radiation injury in patients with recurrent contrast-enhancing intracranial lesions. In these le- sions, the Cho/NAA and Cho/Cr ratios may be the best numeric discriminators. ontrast-enhancing lesions that arise on routine follow-up brain MRI at the site of a previously identified and treated intracranial neoplasm present a significant diagnostic di- lemma. The regions have been subjected to radiation, with or without chemotherapy, and, in some instances, to surgical resection. MRI cannot reliably discriminate tumor recurrence or progression from the inflammatory or ne- crotic changes resulting from radiation [1], al- though the latter can be associated with more specific patterns of enhancement, such as “soap bubble” or “Swiss cheese” enhance- ment [2]. The two entities may be distin- guished by a brain biopsy, the patient’s clini- cal course, or follow-up imaging. Among the noninvasive methods that are available for di- agnosing intracranial tumors, which include SPECT, PET, and diffusion- and perfusion- weighted MRI [3–6], it is mainly proton MR spectroscopic imaging (MR spectroscopy) that has been used in attempts to differentiate tumor from radiation necrosis. Single-voxel MR spectroscopy used in ear- lier investigations [7–10] resulted in interpreta- tive difficulties, with overlapping metabolic ra- tios as a result of partial volume contamination in these histologically heterogeneous lesions. Multivoxel spectroscopic imaging—either 2D or 3D CSI—enables, during the same acquisi- tion, coverage of a larger volume and investi- gation of multiple regions of the lesion and sur- rounding tissue. One could assume that the smaller the sampling unit, the more specific MR spectroscopy will be in these heteroge- neous lesions. However, lesions in the poste- rior fossa and those supratentorial lesions that lie in close proximity to the ventricular system or skull are difficult to evaluate with 2D CSI because of bulk magnetic susceptibility varia- tions. This problem can be largely overcome by using both outer volume suppression slices and in-field-of-view saturation bands for the C Downloaded from www.ajronline.org by 52.73.204.196 on 05/17/22 from IP address 52.73.204.196. Copyright ARRS. For personal use only; all rights reserved