Êtes-vous un étudiant de l'EPFL à la recherche d'un projet de semestre?
Travaillez avec nous sur des projets en science des données et en visualisation, et déployez votre projet sous forme d'application sur Graph Search.
Early detection and localisation of bronchial cancer remains a challenging task. One approach is to exploit the changes in the autofluorescence characteristics of the bronchial tissue as a diagnostic tool with improved sensitivity. Evidence exists that this native fluorescence or autofluorescence of bronchial tissues changes when they turn dysplastic and to carcinoma in situ. There is agreement in the literature that the lesions display a decrease in autofluorescence in the green region of the spectrum under illumination with violet light and a relative increase in the red region of the spectrum is often reported. Imaging devices rely on this principle to detect early cancerous lesions in the bronchi. Based on a previous spectroscopic study, an industrial imaging prototype has been developed to detect early cancerous lesions in collaboration with the firm 'Richard Wolf Endoskope GmbH'. A preliminary clinical trial, involving 20 patients, with this spectrally optimised system proved that autofluorescence can detect lesions that would otherwise have remained invisible even to an experienced endoscopist under white light illumination. A systematic analysis of the autofluorescence images indicated that real-time decisional functions can be defined in order to reduce the number of false positive results. Using this method, a Positive Predictive Value (PPV) of 75% was achieved using autofluorescence only. A PPV of even 100% was obtained when white light mode and autofluorescence mode were combined under the applied conditions. Furthermore, the sensitivity was estimated to be twice as high in AF mode than in WL mode.
David Atienza Alonso, Tomas Teijeiro Campo, Lara Orlandic, Jonathan Dan, Jérôme Paul Rémy Thevenot
Robert West, Andreas Oliver Spitz, Ahmad Abu-Akel