Fourier transform infrared spectromicroscopy and hierarchical cluster analysis of human meningiomas

Limitations of conventional light microscopy in pathological diagnosis of brain tumors include subjective bias in interpretation and discordance of nomenclature. A study using mid-infrared (IR) spectromicroscopy was undertaken to determine whether meningiomas, a group of brain tumors prone to recurrence, could be identified by the unique spectral 'fingerprints' of their chemical composition. Paired, thin (5-μm) cryosections of snap-frozen human meningioma tumor samples removed at elective surgery were mounted on glass (hematoxylin and eosin-stained tissue section) and infrared (unstained tissue section) reflectance slides, respectively. Concordance of the tumor-bearing areas identified in the stained section by a pathologist with the unstained IR tissue section was ensured using a novel digital grid and tumor-mapping system developed in our laboratory. Compared with the normal control, tumor samples from four meningioma patients revealed a marked decrease in bands associated with unsaturated fatty acids, particularly in the bands at 3010, 2920, 2850, and 1735 cm -1 . Spectral datasets were subjected to hierarchical cluster analyses (HCA) using Ward's algorithm for comparison and grouping of similar data groups, and were converted into color-coded digital maps for matching spectra with their respective clusters. False color images of 5 and 6 clusters obtained by HCA identified dominant clusters corresponding to tumor tissue. Corroboration of these findings in a larger number of meningiomas may allow for more precise identification of these and other types of brain tumors.