Optical techniques for diagnosis of cervical precancers: a comparison of Raman and fluorescence spectroscopies

In vivo fluorescence measurements were followed by in vitro NIR Raman measurements on 36 human cervical biopsies, 18 with suspected cervical dysplasia. Fluorescence spectra collected at 337 and 460 nm excitation were used to develop a diagnostic algorithm to differentiate between normal and dysplastic tissues. Using a fluorescence algorithm optimized on another much larger data set, a sensitivity and specificity of 80% and 67% were observed for differentiating squamous intraepithelial lesions (SILs) from all other tissues for the 36 samples. In general, inflammation and metaplasia samples were misclassified at SILs in the larger sample set. A sensitivity and specificity of 57% and 100% were achieved for differentiating high grade from low grade SILs using a second fluorescence algorithm developed with a larger sample set. Using Raman scattering peaks observed at 626 and 1070 cm^-1, SILs could be separated from other tissues with a sensitivity and specificity of 88% and 100% in an algorithm optimized on the data set consisting of the 36 samples from this study. In addition, inflammation and metaplasia samples are correctly separated from the SILs. The band at 1656 cm^-1 could differentiate both SILs from all other tissues as well as high grade from low grade lesions with a sensitivity of 91% and 88% and specificity of 86% and 100%, again using an algorithm optimized on the set of 36 samples. However, these encouraging results must be considered preliminary because of the small sample size. Performance of the fluorescence algorithm to differentiate SILs from squamous normals (sensitivity = 84%, specificity = 82%) was higher than was observed with a smaller data set when tested retrospectively on a larger prediction set of 237 samples.