Raman spectroscopy for early detection of laryngeal malignancy: preliminary results

Abstract

Objective: Raman spectroscopy, the analysis of scattered photons after monochromatic laser excitation, is well established in nonbiological sciences. Recently this method has been used to differentiate premalignant and malignant lesions from normal tissue. Its application for early diagnosis has been explored in a variety of sites (e.g., esophagus, cervix), but not, to date, in laryngeal cancer. The objective of this study was to perform a feasibility study of the use of Raman spectroscopy for early diagnosis of laryngeal malignancy.

Methods: Biopsy specimens were snap-frozen, and top sections were sent for histopathological analysis. Only homogenous samples with clearly defined pathological findings were used in this study: seven histologically normal samples, four exhibiting dysplasia, and four with carcinoma. Samples were defrosted and placed under a Renishaw (Wotton-Under-Edge, UK) System 1000 Raman microspectrometer for analysis. Between 5 and 12 spectra were acquired from each sample, with an excitation wavelength of 830 nm. Average characteristic spectra for each disease or condition were compared. Further multivariate statistical analysis of the data was carried out to evaluate and maximize the differences in the spectra for each disease or condition.

Results: The most visible differences in the spectra occur between 850 and 950 cm(-1) and 1,200 and 1,350 cm(-1). The later peaks are directly related to protein conformation and C-H bond stretch in nucleic acid bases. The relative intensity of the nucleic acid peak increases with progression to malignancy. Use of linear discriminant analysis made it possible to separate the spectra with disease to a greater degree of accuracy than using empirical peak ratio methods alone. Classification results obtained from cross-validation of the discriminant model showed prediction sensitivities of 83%, 76%, and 92% and specificities of 94%, 91%, and 90% for normal, dysplastic, and squamous cell carcinoma of the larynx, respectively.

Conclusions: There was strong evidence to support spectral identification of malignancy and earlier abnormal changes. More substantive studies of the spectral differences between malignant and non-neoplastic tissue are warranted. Raman spectroscopy may become a useful adjunct to pathological diagnosis allowing directed or guided biopsies and assessment of adequacy of resection margins.