Classification of clinical autofluorescence spectra of oral leukoplakia using an artificial neural network: a pilot study

Abstract

The performance of an artificial neural network was evaluated as an alternative classification technique of autofluorescence spectra of oral leukoplakia, which may reflect the grade of tissue dysplasia. Twenty-two visible lesions of 21 patients suffering from oral leukoplakia and six locations on normal oral mucosa of volunteers were investigated with autofluorescence spectroscopy (420 nm excitation, 465-650 nm emission). Pre-scaled spectra were combined with the corresponding visual and histopathological classifications in order to train artificial neural networks. A trained network is mapping input spectra to tissue characteristics, which was evaluated using a blind set of spectra. Abnormal tissue could be distinguished from normal tissue by a neural network with a sensitivity of 86% and a specificity of 100%. Also, classifying either homogeneous or non-homogeneous tissue performed reasonably well. Weak or no correlation existed between spectral patterns and verrucous or erosive tissue or the grade of dysplasia, hyperplasia and hyperkeratosis.