A Multivariate Computational Approach With Hybrid Graphene Oxide Sensor Array for Partial Fulfillment of Breath Acetone Sensing

Abstract

Breath analysis is becoming more prominent in the field of medical science because of its noninvasive nature. In the human breath, there exist thousands of volatile organic components; however, only a limited number are correlated with a specific disease. Acetone is the disease marker for diabetes, and an acetone concentration of more than 1800 ppb in a person’s breath indicates diabetes. In this study, different concentrations of acetone were detected in a dummy exhaled human breath consisting of a large number of interfering volatile organic compounds (VOCs), relative humidity (70%), and synthetic air. An array of six sensors comprised of hybridized graphene oxide (GO) field-effect transistors (FETs) was fabricated and tested for the detection of acetone within the concentration range of approximately 400 ppb to 80 ppm. A GO channel was hybridized with noble metal nanoparticles (NPs) like Au, Pd, and Pt along with WO3 flower-like nanostructures to enhance the acetone selective behavior under suitable back gate voltage in the FET structured sensors. A wide variety of feature vectors were extracted from the sensor data and used via linear discriminant analysis (LDA) and principal component analysis (PCA) to discriminate acetone with different stages of interfaces. Then, the exact acetone concentrations were quantified with a polynomial curve fitting in the sensitivity versus concentration characteristics of all the individual sensors. Acetone concentration in a complex VOC mixture was detected with an average 6.65% relative error by implementing the above technique.