Characterization of CNS Lesions

Sixty-six samples of CSF from 66 patients with a variety of diseases, including tumors, arteriovenous malformations, aneurysms, brain infarctions, and lumbar back pain, were studied with 'H MR spectroscopy at 360 MHz. 'H MR spectroscopy offers a simple means to obtain fast information about different metabolites simultaneously. As compared with the control group, which consisted of 19 CSF samples from the same group of 66 patients, but from individuals who had no abnormal findings on neurologic examination, common clinicochemical tests, or neuroradiologic studies, our preliminary results suggest that tumors and hemorrhages may be differentiated by 'H MR spectroscopy. MR peak intensities relative to lactate peak intensity were used as variables in a statistical analysis to determine the significance of individual resonance intensities in predicting CNS abnormalities. The most important factors for diagnosis were analyzed by means of a multivariate general linear hypothesis and a principal component method. The most important factors for predicting CNS abnormalities in the studied diseases were creatinine, glucose, creatine, citrate, protein content, glutamine, amount of cells, and valine. By using this model for discriminant analysis, we could predict hemorrhages correctly in 88% of cases and tumors in 75% of cases. All samples of controls were determined correctly. In cases of brain infarctions, different signals were observed, which may lead to further characterization of such lesions. 'H MR spectroscopy may offer a simple means for further characterizing CNS lesions. However, this needs confirmation by a prospective study, which would include a larger patient population with different diseases.