Method development: Raman spectroscopy‐based histopathology of oral mucosa

Abstract

An earlier and more accurate detection of (small) cancerous and precancerous lesions in the oral cavity is essential to improve the prognosis of oral squamous cell carcinomas. Raman spectroscopy is being pursued as a potential method to realize this improvement, since the technique provides objective information on a biochemical level and can be used for real‐time guidance of the diagnostic procedure. Since oral mucosal tissue is inhomogeneous and comprises different layers and histological structures, a good understanding of the signal contributions of the individual layers and structures is required for an accurate interpretation of in vivo tissue spectra measurement volumes. The aim of this study was to create a standardized method to collect and analyse the spectral contributions of individual histopathological structures in oral mucosa. The method is based on Raman microspectroscopic mapping of unstained frozen tissue sections and subsequent histopathological annotation of the features in the resulting Raman images. The obtained annotated reference spectra were used as input in an unsupervised hierarchical cluster analysis in order to determine the spectral characteristics and variance within one histo(patho)logical structure. The described method resulted in an annotated database of Raman spectral characteristics of individual histopathological structures encountered in oral tissue. This database can be used as input for the development of classification and quantification algorithms, in order to achieve a high specificity and sensitivity for clinical diagnostic instruments. Additionally, this database can be used to optimize the exact location and measurement volume of in vivo measurements. Copyright © 2013 John Wiley & Sons, Ltd.

Investigation of the potential of Raman spectroscopy for oral cancer detection in surgical margins

“…25 Briefly, for each individual mapping experiment, the Raman data were processed and analyzed with software developed in-house that operates in a MATLAB environment (MATLAB 7.5.0 (R2007b), MathWorks, MA, USA) and uses the multivariate toolbox PLS-toolbox 7.0.0c (EigenVector Research, WA, USA). After removing cosmic ray events, the raw data were calibrated.…”

“…25 The contributions of carotenoid and hydroxyapatite represented biological signals that were considered non-informative, since these signals showed a large influence on the scaling procedure, resulting in noninformative differences between similar histopathological structures in different Raman experiments. Finally, the wavenumber range of the spectra that were included in the data-analysis was restricted to the 400-1800/cm.…”

“…25 Briefly, principal components analysis (PCA) was used to reduce the data set without significant loss of information. 27 All data were mean-centered before PCA.…”

“…For each experiment, the optimal number of K-means clusters was defined, based on the best match between histo(patho)logical structures that were present in the mapped region and the pseudo-color Raman image, as described in detail earlier. 25 Figure 2 shows representative examples of H&E-stained sections and their corresponding pseudo-color Raman images ( Figure 2). The pathologist was blinded for the corresponding spectra.…”

“…In previous research, we developed a method of collecting detailed spectral information of individual histo(patho)logical structures based on in vitro Raman mapping experiments. 25 In the current study, we used that method to create a spectral database that comprises OCSCC and healthy tissue structures of the tongue. We investigated how well the tissue spectra enable discrimination between OCSCC and surrounding healthy tissue structures.…”