Improving endoscopic detection of lung cancer using autofluorescence spectroscopy analysed by a neural network (RNN.5316)

Lung cancer is by far the commonest cancer in the Western world and the incidence is still rising. It has a very high mortality rate due to the late stage at which lung tumours are usually diagnosed. Over the last decade new endoscopic techniques have been developed for screening of high-risk groups in order to be able to detect these tumours earlier in their development and improve the outlook for these patients. Endoscopic autofluorescence imaging, introduced a few years a go, showed a very high sensitivity. Regretfully, the clinical success was compromised by a very low specificity (i.e. too many false positives). Spectroscopy of tissue autofluorescence combined with classification by a neural net has shown to be a more powerful technique. Aim of the present project is to improve endoscopic detection of lung cancer by combining the high sensitivity of endoscopic fluorescence imaging with the excellent specificity of autofluorescence spectroscopy analysed by neural network. So, we detect lesions with fluorescence imaging, obtain spectroscopic information from each lesion and let a neural net select the malignant ones. The first technological challenge will be to develop an endoscopic device capable of autofluorescence imaging as well as spectroscopy in a (central) spot. For this we established the close collaboration of an industrial partner (Karl Storz & Co GmbH). A prototype will be constructed (based on an existing fluorescence endoscope), which will be optimised after a pilot study. With the improved prototype we will acquire a large series of clinical data from patients having or suspected of having lung cancer as well as from other patients coming to the hospital for routine bronchoscopy. The second challenge will be to develop and train a neural network that is capable of real time differentiation cancer and benign lesions on the basis of autofluorescence spectra. The preliminary results in the oral cavity look very promising and have induced great interest. In fact, an industrial partner has shown genuine interest to participate in developing this in particular for the lungs.

Er zijn nog geen resultaten bekend.

A company, a research institute and a hospital are involved in this project.