Combined texture features for improved classification of suspicious areas in autofluorescence bronchoscopy

Autofluorescence bronchoscopy (AFB) has been utilized over the past decade, proving to be a powerful tool for the detection and localization of premalignant and malignant lesions of the airways. AFB is, however, characterized by low specificity and a high rate of false positive findings (FPFs). The majority of FPFs are due to inflammations, as they often fluoresce at the same wavelengths with cancer. According to several clinical trials, the percentage of the FPFs is about 30%. In this paper we present an intelligent computing system based on combined texture features, feature selection methods and classification models, for improved classification of suspicious areas of the bronchial mucosa, in order to decrease the rate of FPFs, to increase the specificity and sensitivity of AFB and enhance the overall diagnostic value of the AFB method.