A Prototype of Graphene E-Nose for Exhaled Breath Detection and Label-Free Diagnosis of Helicobacter Pylori Infection

Abstract

Helicobacter pylori (HP), a common microanaerobic bacteria that lives in the human mouth and stomach, is reported to infect ≈50% of the global population. The current diagnostic methods for HP are either invasive, time-consuming, or harmful. Therefore, a noninvasive and label-free HP diagnostic method needs to be developed urgently. Herein, reduced graphene oxide (rGO) is composited with different metal-based materials to construct a graphene-based electronic nose (e-nose), which exhibits excellent sensitivity and cross-reactive response to several gases in exhaled breath (EB). Principal component analysis (PCA) shows that four typical types of gases in EB can be well discriminated. Additionally, the potential of the e-nose in label-free detection of HP infection is demonstrated through the measurement and analysis of EB samples. Furthermore, a prototype of an e-nose device is designed and constructed for automatic EB detection and HP diagnosis. The accuracy of the prototype machine integrated with the graphene-based e-nose can reach 92% and 91% in the training and validation sets, respectively. These results demonstrate that the highly sensitive graphene-based e-nose has great potential for the label-free diagnosis of HP and may become a novel tool for non-invasive disease screening and diagnosis.

Keywords: Helicobacter pylori; exhaled breath diagnosis; e‐nose prototype; gas sensors; graphene oxide.

Figure 1

SEM pictures of different materials. a‐1) rGO (insert in (a‐1)) and rGO‐PDDA; a‐2) MoS₂; a‐3) AuNPs; a‐4) AgNPs; a‐5) rGO‐PDDA‐Co/Fe; a‐6) rGO‐PDDA‐Co/Cu; a‐7) rGO‐PDDA‐Au/Ag; a‐8) rGO‐PDDA‐Ag; a‐9) rGO‐PDDA‐MoS₂; a‐10) rGO‐PDDA‐Ce; a‐11) rGO‐PDDA‐Fe; a‐12) rGO‐PDDA‐Co. The UV–vis extinction spectrum of b) AuNPs; c) AgNPs; d) The Raman spectra of different materials; e,f) The XPS spectra of different materials. g) The 8‐channel sensing array chip. h) The pictures of I–V curves of different materials on electrodes; i) The baseline noises of different materials on electrodes.

Figure 2

The radar of response values of rGO‐PDDA‐M sensor; a) different composites; b) different gases; Response curves of rGO‐PDDA‐M to c) 0.1‐1 ppm NO; d) 2–15 ppm isoprene; e) Response repeatability of rGO‐PDDA‐M to 10 ppm isoprene. Langmuir fitting of rGO‐PDDA‐M toward f) NO; g) Linear fitting of isoprene; h) Response dynamic curves of rGO‐PDDA‐M to 1 ppm NO (The blue highlight was t₅₀ of each sensing layer); i) The t₅₀ values of rGO‐PDDA‐M to 1 ppm NO; j) PCA analysis conducted based on the normalized Re% and t₅₀ of four typical gases.

Figure 3

Response dynamic curves of rGO‐PDDA‐M sensor array toward EB sample of a) pure EB; b) spiked EB (1 ppm NH₃). c) 2D OPLS‐DA; d) PCA e) 3D OPLS‐DA and f) PCA results according to Re% of rGO‐PDDA‐M sensor array toward pure EB and spiked EB (1 ppm NH₃).

Figure 4

a) Box charts of response values of clinical EB samples; (a‐1‐a‐8) Response values of rGO‐PDDA‐Co; rGO‐PDDA‐Fe; rGO‐PDDA‐Ce; rGO‐PDDA‐MoS₂; rGO‐PDDA‐Ag; rGO‐PDDA‐Au/Ag; rGO‐PDDA‐Co/Cu; and rGO‐PDDA‐Co/Fe toward EB samples of healthy person (colored) and people infected with HP (black); b) Radar map of VIP values of Re% variable of clinical EB samples analysis at 8 sensing layers of the sensor array; c) The OPLS‐DA result of clinical EB samples analysis at 4 sensing layers of the sensor array. (n = 204; α = 0.05; P < 0.05; n is sample size; α is the level of significance test; and the P value is the probability that if the null hypothesis is true, an outcome more extreme than the obtained sample observations will occur.).

Figure 5

a) The structure diagram of the microsensor array; b) The real picture of the microsensor array; c) The schematic of the e‐nose prototype; d) The working diagram of the e‐nose prototype; e) The photo of the actual e‐nose prototype; The dimensions of the e‐nose prototype were determined to be 22 cm × 36 cm × 26 cm.

Figure 6

Discriminant results of validation set of EB samples based on Lasso regression model. a) median normalization; b) sum normalization; c) max‐min normalization. Lasso model analysis results of d) training set; e) validation set; and f) external test set of 225 cases EB samples. ROC curves analysis of g) training set; h) validation set; and i) external test set of 225 cases EB samples.