Fourier-transform infrared (FTIR) microspectroscopy, combined with Pri
ncipal Component Analysis (PCA), was applied in the study of exfoliate
d cervical cells from 272 patients. Six spectra were recorded for each
patient, and these were visually sorted into two types (type 1 and ty
pe 2), based on their profiles. Spectra designated type 1 exhibited a
profile characteristic of normal epithelial cells, with intense glycog
en bands at 1022 cm(-1) and 1150 cm(-1), and a pronounced symmetric ph
osphate stretch at 1078 cm(-1). Spectra designated type 2 exhibited fe
atures suggestive of dysplastic or malignant transformation, with pron
ounced symmetric and asymmetric phosphate modes and a reduction in gly
cogen-band intensity. Of the 272 patients, 68.6% of samples exhibited
only type 1 profiles for all six recorded spectra, 29.4% of samples yi
elded at least one type 2 spectrum in any of the six recorded spectra
and 2% of samples were inconclusive. Of the 68.6%, 86% were diagnosed
normal by Pap smear with no follow up biopsy ordered, 7% were diagnose
d abnormal by biopsy, 5% normal by biopsy and 2% were still inconclusi
ve. For the remaining 29.4% of classified samples, 71% had shown an ab
normal Pap result. These 71% were subsequently biopsied, and 87% were
confirmed abnormal. The association of type 2 spectra and abnormality
was further corroborated by spectra of cultured malignant cells from t
he HeLa cell line that displayed a profile similar to type 2 spectra i
n the 1300-950 cm(-1) region. PCA decomposition using a reduced data m
atrix resulted in a score plot that showed general separation of the v
isually categorised spectra. This study demonstrates the potential of
automated FTIR cervical screening technology in the clinical environme
nt. (C) 1996 John Wiley & Sons, Inc.