Objective: Raman spectroscopy, the analysis of scattered photons after mono
chromatic laser excitation, is well established in nonbiological sciences.
Recently this method has been used to differentiate premalignant and malign
ant lesions from normal tissue. Its application for early diagnosis has bee
n explored in a variety of sites (e.g., esophagus, cervix), but not, to dat
e, in laryngeal cancer. The objective of this study was to perform a feasib
ility study of the use of Raman spectroscopy for early diagnosis of larynge
al malignancy. Methods: Biopsy specimens were snap-frozen, and top sections
were sent for histopathological analysis, Only homogenous samples with cle
arly defined pathological findings were used in this study: set-en histolog
ically 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, Av
erage characteristic spectra far each disease or condition were compared. F
urther multivariate statistical analysis of the data was carried out to eva
luate and maximize the differences in the spectra for each disease or condi
tion. Results: The most visible differences in the spectra occur between 85
0 and 950 cm(-1) and 1200 and 1350 cm(-1). The later peaks are directly rel
ated to protein conformation and C-H bond stretch in nucleic acid bases. Th
e relative intensity of the nucleic acid peak increases with progression to
malignancy. Use of Linear discriminant analysis made it possible to separa
te the spectra with disease to a greater degree of accuracy than using empi
rical 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, dysplasti
c, and squamous cell carcinoma of the larynx, respectively, Conclusions: Th
ere was strong evidence to support spectral identification of malignancy an
d earlier abnormal changes. More substantive studies of the spectral differ
ences between malignant and non-neoplastic tissue are warranted. Raman spec
troscopy may become a useful adjunct to pathological diagnosis allowing dir
ected or guided biopsies and assessment of adequacy of resection margins.