Discriminating Artificial Cancer Breath Using an Electronic Nose: K-Nearest Neighbors Versus Long-Short Term Memory Network

This paper presents the use of k-NN for the classification of healthy human breath with or without the addition of lung cancer biomarkers. 236 breath samples collected from 17 persons over four months were analyzed by a custom electronic nose using commercial metal oxide sensors. About 90% of the samples were correctly classed by the model. Long-Short Term Memory Neural Network could show promising results in this task as well and are under investigation.