FTIR MICROSPECTROSCOPIC STUDY OF CELL-TYPES AND POTENTIAL CONFOUNDINGVARIABLES IN SCREENING FOR CERVICAL MALIGNANCIES

FTIR microscopy was applied to the analysis of cell types and other va riables present in Pap smears to ascertain the limitations of infrared spectroscopy in the diagnosis of cervical cancer and dysplasia. It wa s found that leukocytes, and in particular lymphocytes, have spectral features in the phophodiester region (1300-900 cm(-1)) suggestive of w hat has previously been described as changes indicative of malignancy. Endocervical cells and fibroblasts have similar spectral features to HeLa cells and consequently could also confound diagnosis. The use of ethanol as a fixative and dehydrating agent results in retention of gl ycogen in cervical cell types and thus minimizes spectral changes in t he glycogen region due to sampling technique. Spectra of seminal fluid s exhibit strong bands in the phosphodiester/carbohydrate region; howe ver, sperm contamination should be easily detectable by the presence o f a distinctive doublet at 981/968 cm(-1). Erythrocyte spectra exhibit a reduction in glycogen band intensity, but can be discerned by a rel atively low-intensity upsilon(s) PO2- band. Endocervical mucin spectra exhibit a reduction in glycogen bands and a very pronounced upsilon(s ) PO2- band, which is similar in intensity to the corresponding band i n HeLa cells. Thrombocytes have strong bands in the phosphodiester reg ion, but thrombocytes can be discerned from other cell types by the pr esence of two small broad bands at 980 and 935 cm(-1). Candida albican s is characterized by strong bands in the polysaccharide region which could potentially obscure diagnostic bands if C. albicans is present i n large numbers. Spectra of bacteria common to the female genital trac t, in general, also have strong absorptions in the polysaccharide regi on; however, bacterial contamination is usually minimal and would not be expected to obscure cervical cell spectra. Nylon threads and bristl es from cervical sampling implements produce characteristic IR profile s which allow for easy identification. Given the number of potential c onfounding variables associated with cervical cytology, a multivariate statistical or neural network analysis would appear to be necessary b efore the implementation of FTIR technology in clinical laboratories. (C) 1998 John Wiley & Sons, Inc.