INFRARED MICROSPECTROSCOPY AND ARTIFICIAL NEURAL NETWORKS IN THE DIAGNOSIS OF CERVICAL-CANCER

Infrared spectra of 88 normal and 32 abnormal (mild to severe dysplasi a) cervical smear samples were used as a databank to investigate the u sefulness of artificial neural networks (ANN) in the diagnosis of cerv ical smears. The spectra were first reduced, using principal component analysis (PCA), to seven wavenumber components that are the major con tributors to the variance. A number of different ANN architectures wer e investigated that could differentiate between normal and abnormal ce rvical smears. Although the ANNs were trained to differentiate only no rmal from abnormal smears, the results using an independent test data set indicated that within the abnormal category mild dysplasia could b e distinguished from severe dysplasia. The results using this restrict ed data set indicate that neural networks coupled to infrared microspe ctroscopy could provide an alternative automated means of screening fo r cervical cancer.