Neural-network-driven Electronic Nose Enhancing Artificial Olfaction in Non-invasive Diagnostics

This paper presents an e-nose specifically designed for non-invasive diagnostics and human volatilome analysis. The sensing technology is based on a 10-sensors array of both commercial Metal Oxide (MOX) gas sensors and custom-fabricated counterparts. Thanks to a versatile pneumatic system, it is capable of analyzing response signals from various sample types, including exhaled breath and the headspace of human biological samples. A neural-network-based model is adopted to enhance the classification capability. The device's effectiveness is demonstrated through experimental tests with both chemical standards and mixtures resembling human biosamples, achieving a 97.1% classification accuracy with 7 prepared test samples. The experimental results, along with the capability to discriminate correctly the test samples in presence of water, confirm the system's efficacy in the context of non-invasive diagnostics and human volatilome analysis.