AbstractAutofluorescence bronchoscopy (AFB) is currently considered the most promising diagnosis modality for the detection of early stage lung cancer with a sensitivity which normally exceeds 90%, although its specificity is rather low. The number of false positive results can be reduced by the combination of autofluorescence (AF) imaging with quantitative spectroscopy. The aim of this study was to reveal additional spectral characteristics of AF emission which may have the potential to maximize the endogenous contrast between normal mucosa and bronchial malignancyAF emission spectra were recorded at 532 nm laser excitation in the course of AFB of totally 20 patients with central-type X-ray negative lung cancer. According to the results of histopathological analysis, two sets were made up of the spectral data corresponding to normal bronchial epithelium and malignant lesions.Comparative studies of nine basic spectral characteristics ofReal time measurements and estimation of all important spectral information on the basis of the probabilistic approach is assumed to be useful for real time recommendations for biopsies and for minimizing the number of false positive results during the course of AFB. For the realization of this approach, further trials are necessary to collect a large number of AF spectra samples, corresponding to different tissue pathologies such as inflammation, metaplasia, and dysplasia.
A semantic-based probabilistic approach for real-time video event recognition
