Rationale and objectives: To compare the ability of normalized versus non-normalized metabolite ratios to differentiate recurrent brain tumor from radiation injury using magnetic resonance spectroscopy (MRS) in previously treated patients.
Materials and methods: Twenty-five patients with previous diagnosis of primary intracranial neoplasm confirmed with biopsy/resection, previously treated with radiation therapy (range, 54-70 Gy) with or without chemotherapy and new contrast enhancing lesion on a 1.5 T magnetic resonance imaging at the site of the primary neoplasm participated in this retrospective study. After MRS, clinical, radiological, and histopathology data were used to classify new contrast-enhancing lesions as either recurrent neoplasm or radiation injury. Volume of interest included both the lesion and normal-appearing brain on the contralateral side. Non-normalized metabolic ratios were calculated from choline (Cho), creatine (Cr), and N-acetylaspartate (NAA) spectroscopic values obtained within the contrast-enhancing lesion: Cho/Cr, NAA/Cr, and Cho/NAA. Normalized ratios were calculated using the metabolic values from the contralateral normal side: Cho/normal creatinine (nCr), Cho/normal N-acetylaspartate (nNAA), Cho/normal choline, NAA/nNAA, NAA/nCr, and Cr/nCr. Results were correlated with the final diagnosis by Wilcoxon rank-sum analysis.
Results: Two of three non-normalized ratios, Cho/NAA (sensitivity 86%, specificity 90%) and NAA/Cr (sensitivity 93%, specificity 70%) significantly associated with tumor recurrence even after correcting for multiple comparisons. Of the six normalized ratios, only Cho/nNAA significantly correlated with tumor recurrence (sensitivity 73%, specificity 40%), but did not remain significant after correcting for multiple comparisons.
Conclusion: Cho/NAA and NAA/Cr were the two ratios with the best discriminating ability and both had better discriminating ability than their corresponding normalized ratios (Area under the curve = 0.92 versus 0.77, AUC= 0.85 vs. 0.66), respectively.
Copyright © 2011 AUR. Published by Elsevier Inc. All rights reserved.