Cutoff value of choline concentration reliably reveals high-grade brain tumors among other contrast-enhancing brain lesions.

Background and Aim  To evaluate whether there is a cutoff value for a metabolite concentration measured by 1 H MR spectroscopy (MRS), which can be used to differentiate malignant brain tumors (high-grade gliomas, primary CNS lymphomas [PCNSL] and metastases) from other contrast-enhancing lesions like low-grade gliomas and non-neoplastic lesions. Material and Methods  1 H MRS was performed in 252 consecutive patients with space-occupying brain lesions which were enhanced with application of a contrast agent. Concentrations of N-acetyl-aspartate, total creatine, choline containing metabolites (total choline, tCho), lipids, and lactate were evaluated from the contrast-enhancing part of the lesions and from the normal appearing brain tissue. Linear discriminant analysis was used to find the best predictor for malignant brain tumors. In addition, receiver operating characteristic analysis (ROC) was performed to determine a cutoff value for the best predictor in detecting malignant brain tumors with a specificity of >95%. Results  All brain tumors and 20 out of 47 nonneoplastic lesions were examined histopathologically. The remaining 27 diagnoses were based on MR imaging, clinical findings, and follow-up. The final diagnosis was 134 high-grade gliomas (WHO grade III/IV), 36 metastases, 9 PCNSL, 8 low-grade gliomas (WHO grade I/II), 34 infections, 9 infarctions, 2 hematomas, and 2 vasculitides. 18 patients were excluded due to insufficient spectral quality. The tCho concentration was the best predictor to differentiate malignant brain tumors from enhancing low-grade gliomas or non-neoplastic lesions (F=26.6 [df: 25.833], p Conclusion  1 H MRS reliably differentiates malignant brain tumors from other contrast-enhancing brain lesions. At least a 40% increase of tCho compared to normal brain tissue indicates a malignant tumor (WHO grade III/IV gliomas, PCNSL, metastases) with >90% specificity and sensitivity.