Determination of histopathological tumor grade in neuroepithelial brain tumors by using spectral pattern analysis of in vivo spectroscopic data

Object. In this study, H-1 magnetic resonance (MR) spectroscopy was prospectively tested as a reliable method for presurgical grading of neuroepithelial brain tumors. Methods. Using a database of tumor spectra obtained in patients with histologically confirmed diagnoses, 94 consecutive untreated patients were studied using single-voxel H-1 spectroscopy (point-resolved spectroscopy; TE 135 msec, TR 1500 msec). A total of 90 tumor spectra obtained in patients with diagnostic H-1 MR spectroscopy examinations were analyzed using commercially available software (MRUI/VARPRO) and classified using linear discriminant analysis as World Health Organization (WHO) Grade 1111, WHO Grade 111, or WHO Grade IV lesions. In all cases, the classification results were matched with histopathological diagnoses that were made according to the WHO classification criteria after serial stereotactic biopsy procedures or open surgery. Histopathological studies revealed 30 Grade I/II tumors, 29 Grade III tumors, and 31 Grade IV tumors. The reliability of the histological diagnoses was validated considering a minimum postsurgical follow-up period of 12 months (range 12-37 months). Classifications based on spectroscopic data yielded 31 tumors in Grade I/II, 32 in Grade 111, and 27 in Grade IV. Incorrect classifications included two Grade 11 tumors, one of which was identified as Grade III and one as Grade IV; two Grade III tumors identified as Grade 11; two Grade III lesions identified as Grade IV; and six Grade IV tumors identified as Grade III. Furthermore, one glioblastoma (WHO Grade IV) was classified as WHO Grade I/II. This represents an overall success rate of 86%, and a 95% success rate in differentiating low-grade from high-grade tumors. Conclusions. The authors conclude that in vivo H-1 MR spectroscopy is a reliable technique for grading neuroepithelial brain tumors.