Criticality of human breath detection with a portable device II: data analysis

Human breath is largely composed of oxygen, carbon dioxide, water vapor, nitric oxide, and numerous volatile organic compounds (VOCs) [1, 2]. Changes in the concentration of the molecules in VOCs could suggest various diseases or at least changes in the metabolism. Indeed, breath gases are recognized to be excellent indicators of the presence of diseases and clinical conditions. Such gases have been identified as biomarkers using accurate but expensive benchtop instrumentations such as gas chromatography (GC) or electronic nose (e-nose) [1]. As a consequence, in recent years, it has been stimulated the necessity to develop a portable device for breath analysis, easy to use, and feasible for patients living far from medical structures or physicians. In the framework of SEMEOTICONS (SEMEiotic Oriented Technology for Idividual's CardiOmetabolic risk self-assessmeNt and Self-monitoring) European Project, we developed a low cost, portable, easy-to- use device for the analysis of breath composition: the Wize Sniffer (WS). The WS captures breath samples, the chemical selective sensors sense the sample and accordingly form a sort of odor-print of healthy people or patients with known and specific diseases, in order to evaluate the well-being state of a human subject [3, 4]. It should be noted that does not exist a general definition of "well-being state", rather some indices for well-being that can be correlated to cardio-metabolic risk, which is representing the leading cause of worldwide mortality [3, 5]. The first prototype of such device is based on commercial, semiconductor-based gas sensors. This type of gas sensors is very robust and easy to integrate. Nevertheless, they are non- selective at all. This leads to several issues for data analysis. In this report we focus our attention on the different strategies for data analysis, evaluating also their performances and outcomes.