MEMS sensor array-based electronic nose for breath analysis—a simulation study

Abstract

The paper presents a simulation study of breath analysis based on theoretical models of microelectromechanical structure (MEMS) cantilever sensor array. The purpose of this study is to suggest a methodology for the development of MEMS electronic nose (e-nose) for monitoring disease-specific volatiles in exhaled breath. Oxidative stress and diabetes are taken as case studies for the assessment of e-nose designs. The detection of ethane for general oxidative stress, isoprene for hypoxia, and acetone for diabetes are considered for targeted detection. A number of volatiles concurrently present in the exhaled breath are taken as interferents. The MEMS cantilevers are coated with volatile-selective polymers and are analyzed in both the static and dynamic modes. The sensor array is defined by polymer selections based on three data mining methods: principal component analysis (PCA), fuzzy c-means clustering (FCM), and fuzzy subtractive clustering (FSC). This utilizes vapor/polymer partition coefficients as a database. Analyses are carried out to find optimal combinations of the polymer selection method and cantilever sensing mode. Virtual breath analysis experiments are analyzed by PCA for target discrimination. It is found that no single combination works best in all conditions. The acetone (diabetes) detection is best in both sensing modes with the polymers selected by FSC; the isoprene (hypoxia) is detectable only in static sensing mode with polymers selected by FCM clustering; and the ethane (oxidative stress) detection is possible by all sensing modes and polymer selections, provided the breath samples are preconcentrated. This study suggests that it is difficult to realize a single general-purpose MEMS breath analyzer. The dedicated analyzers for specific disease indications can however be made with an optimal combination of sensing mode and polymer coatings.

A Bio‐Inspired Neuromorphic Sensory System

“…[40][41][42][43][44][45] Among them, formaldehyde, acetone, and toluene are receiving increasing attention for being the crux biomarkers for disease diagnosis and important indicators of indoor air quality. [46][47][48][49] A sensor array of six commercial sensors (TGS2611, TGS2600, TGS2602, TGS2603, TGS2610, TGS2620) were used to detect four kinds of gases (formaldehyde, ethanol, acetone, toluene). During the beginning 200 s, pure air was inputted into the test chamber and the sensor resistances fluctuated around the baseline due to noise.…”

Review of the algorithms used in exhaled breath analysis for the detection of diabetes

“…In addition to using direct classification, a diagnosis based on predicted blood sugar concentration has also been proposed [65]. The researches were carried out the analysis on the breath samples taken from patients [57,58,60,62,64,[66][67][68], and simulated acetone concentrations [40,[68][69][70]. The algorithms presented in the papers have been developed mainly in MATLAB ® programming and numeric computing environment [62,63,71] and in Python programming language with Keras interface for designing artificial neural networks (ANNs) [64].…”

“…with a laser reflectometer or by measuring capacity changes. Another measurement method is the dynamic mode sensing, which involves electrostatic induction of the cantilever by applying AC voltage to the capacitor plates and measuring electronically or by a laser Doppler vibrometer the change in resonance frequency caused by the presence of gases [70,116]. Gupta et al designed a polymer-coated MEMS cantilever sensor.…”

Detection of Liver Dysfunction Using a Wearable Electronic Nose System Based on Semiconductor Metal Oxide Sensors

“…However, unlike other analytical methods, an e-nose does not detect directly specific volatile organic components (VOCs); rather, it builds chemical patterns to form an identity. The sensor array produces output patterns that represent VOCs in the breath (or different substances), and the data processing extracts a set of mathematical descriptors that represent the signature of the breath sample as a pattern [ 15 ]. The detection of the input signal occurs depending on the operating principle implemented in the sensor arrays.…”