An investigation on electronic nose diagnosis of lung cancer.

The use of gas sensor arrays as medical diagnosis instruments has been proposed several years ago. Since then, the idea has been proven for a limited number of diseases. The case of lung cancer is particularly interesting because it is supported by studies that have shown the correlation between the composition of breath and the disease. However, it is known that many other diseases can alter the breath composition, so for lung cancer diagnosis it is necessary not only to detect generic alterations but those specifically consequent to cancer. In this paper an experiment, performed in the bronchoscopy unit of a large hospital, aimed at discriminating between lung cancer, diverse lung diseases and reference controls is illustrated. Results show not only a satisfactory identification rate of lung cancer subjects but also a non-negligible sensitivity to breath modification induced by other affections. Furthermore, the effects of some compounds frequently found in the breath of lung cancer subjects have also been studied. Results indicate that breath samples of control individuals drift towards the lung cancer group when added with either single or mixtures of these alleged cancer-related compounds.