Editorial Board

For accurate diagnosis of diseases, gas sensors with new nanomaterials have been studied. Since the difference in the concentration of the biomarkers in EB is minute, the gas sensors should be highly sensitive.

Recently, semiconductor metal oxide (SMO) gas sensors with new sensing materials have been researched for detecting sub-ppm level of acetone for diabetes diagnosis. Fig. 2(a) illustrates a synthetic scheme of ZnO-CuO core-hollow cube nanostructures for acetone detection [9]. CuO hollow cubes were synthesized by thermal oxidation of CuO₂ cubes and then directly attached to ZnO spherical aggregates. They formed core-shell-like structures and p-n heterojunctions, and hence showed remarkable response (R_gas/R_air, where R_gas is the resistance in target gas, and R_air is the resistance in atmospheric condition) of 11.14 at ppm acetone, as shown in Fig. 2(b). Fig. 2(c) shows responses of ZnO-CuO nanostructures and previously reported sensing materials; the ZnO-CuO core-hollow cube nanostructures exhibit much higher response to acetone. The limit of detection (LOD) of the sensor is estimated as 9 ppb. Hanh et al. fabricated Pt-Zn₂SnO₄ hollow octahedra sensors with hydrothermal reaction and detected acetone at ppb level with high selectivity and repeatability [10]. Ma et al. fabricated a highly sensitive gas sensor using phase transition of h-CoO nanoparticles to CO₃O₄ nanoplates [11]. Thermal oxidation of h-CoO led to deposition of Co₃O₄, and phase-transitioned Co₃O₄ showed high sensitivity and selectivity to acetone. Li et al. synthesized SnO₂ nanosheets using the hydro-solvothermal treatment and decorated them with PdAu bimetallic nanoparticles [12]. The SnO₂-PdAu nanosheet sensors showed high responses to acetone and formaldehyde and their LODs were 45 and 30 ppb, respectively. Fig. 2(d) shows Co₃O₄ nanofibers and Cds nanospheres composite fabricated using electrospinning with a hydro-thermal method [13]. Fig. 2(e) shows dynamic response of CdS/Co₃O₄ sensor to acetone gas under green light illumination. The green light generates electron-hole pairs on the surface of the sensor film, and the photogenerated holes promote reaction between acetone and nanocomposites. Fig. 2(f) shows responses to diabetes breath, healthy breath, and 2 ppm acetone and verifies that the sensor can diagnose diabetic patients. Fig. 2(g) illustrates schematics of the fabrication of Pt@In₂O₃ core-shell nanowires [14]. As shown in Fig. 2(h), the fabricated Pt@In₂O₃ core-shell nanowire sensor exhibited high sensitivity to low concentration of acetone with the detection limit of 10 ppb. Fig. 2(i) shows portable sensing device with Pt@In₂O₃ core-shell nanowire (NW) sensor, which successfully distinguished between healthy and diabetic volunteers in real-time.

Fig. 2. 
Nanosensors for diabetes diagnosis. (a) Schematic of synthesis of ZnO-CuO core-hollow cube nanostructures. (b) Dynamic responses of ZnO-CuO sensors to acetone. (c) Response comparison to previously reported sensors. Reprinted with permission from Ref [9]. Copyrights (2020) American Chemical Society. (d) SEM image of CdS/Co₃O₄ nanocomposite. (e) Dynamic responses of the CdS/Co₃O₄ gas sensor to acetone in the presence or absence of green light. (f) Response results of the sensor tested to simulated diabetes breath, acetone, and healthy breath Reprinted with permission from Ref [13]. Copyrights (2022) Elsevier. (g) Schematic of the electrospinning fabrication of Pt@In₂O₃ core-shell nanowires. (h) Dynamic responses of Pt@In₂O₃ core-shell NW to acetone gas. (i) Schematic of the portable sensing device and response to the exhaled breath from healthy volunteers. Reprinted with permission from Ref [14]. Copyrights (2018) Springer Nature.

Selective gas sensors, for detection of specific VOCs related to lung cancer, have also been investigated. Unlike diabetes diagnosis, gas sensors targeting various biomarker candidates have been studied.

Nam et al. improved gas sensors for the toluene detection using highly porous WO₃ nanofibers (NFs) by combining Pd catalysts [21]. Fig. 3(a) illustrates fabrication process of the Pd@ZnO-WO₃ heterogeneous nanofibers for detection of toluene [22]. Electrospun Pd@ZIF-8 was oxidized to obtain ZnO and the Pd@ZnO particles were embedded onto WO₃ NFs during calcination process. Fig. 3(b) shows the sensor responses to the toluene exposure. The Pd@ZnO-WO₃ NFs exhibited higher responses than other sensing materials. The response time was much shorter (< 20 s) and the LOD was estimated as 100 ppb. Selectivity of the Pd@ZnO-WO₃ was investigated with interfering analyte gases, which showed higher sensitivity to toluene rather than other gases. The Pd-embedded/WO₃ NFs were fabricated by electrospinning and calcination. The Pd functionalized WO₃ nanofibers-based gas sensor exhibited 5.5 response to 1 ppm toluene gas and remarkable selectivity against H₂S gas. Lee et al. synthesized HKUST-1-based CuO decorated with Pt nanoparticles (NPs) using sonochemical method [23]. The CuO/Pt NPs gas sensors showed high response over 82% to formaldehyde in the concentration range of 100–1500 ppb. Majidi and Nadafan analyzed the possibility of detecting biomarkers with γ-graphyne and twin-graphene sheets [24]. Based on density functional theory, they analyzed the effect of adsorption of benzene, styrene, aniline, and o-toluidine. γ-graphyne and twin-graphene sheets can be used as aniline and o-toluidine by detecting the changes in the electric dipole moment and energy band gap.

Fig. 3. 
Nanosensors for lung cancer diagnosis. (a) Schematic of fabrication for the Pd@ZnO-WO₃ nanofibers. (b) Response results of pristine WO₃, ZnO-WO₃, Pd(polyol)-WO₃, Pd(polyol)-ZnO-WO₃, and Pd@ZnO-WO₃ NFs for toluene detection. Reprinted with permission from Ref [22]. Copyrights (2016) American Chemical Society. (c) Images of the micro-analytical units with a single SnO₂ gas sensor for detection of VOCs biomarkers. (d) Normalized conductance results of biomarkers in different mixtures. Reprinted with permission from Ref [25]. Copyrights (2018) Elsevier.

Not only high-performance gas sensors, but also detection devices have been researched [25]. Fig. 3(c) shows the micro-analytical units for the detection of volatile biomarkers. The detecting device includes micro-preconcentrator with a zeolite adsorbent, GC micro-column, and SnO₂ gas sensor. Target gas passes through micro-preconcentrator and GC micro-column, the gas sensor can selectively detect biomarkers in a short time. Fig. 3(d) shows electrical responses to various biomarkers, and the LODs were estimated to be 24±3, 5±1, 21±3, and 2515±5 ppb for toluene, o-xylene, propanol, and cyclohexane, respectively.

Diagnosing diabetes and lung cancer with a single sensor is difficult because of a variety of VOCs related to the diseases and interference of disruptive factors, like temperature and humidity, with sensor signals. Therefore, the VOC patterns in EB have been studied using gas sensor arrays, called as an electronic nose (E-nose). The E-nose system includes gas sensors with different sensing materials and generates response patterns. Various gas sensors have been utilized for the E-nose systems.

Lee et al. fabricated MEMS microheater integrated gas sensor arrays with four different sensing materials based on CuO [26]. The gas sensor arrays detected acetone with high sensitivity and distinguished it from other VOCs using principal component analysis (PCA). Orbe et al. fabricated breath analyzer for monitoring metabolic fat burning as shown in Fig. 4(a) [27]. The acetone in EB and beta-hydroxybutyric acid (BOHB) are correlated biomarkers of the metabolic fat burning by lipolysis in the liver. Fig. 4(b) and (c) show schematic of the breath intake operation. The EB automatically flows to metal oxide gas sensor arrays, and responses to breath samples are collected. Recurrent neural network (RNN) algorithm was used to estimate the values of BOHB using collected response data. Fig. 4(d) shows feature coefficients of the variables provided by the sensors and the The result shows feature impact and feature impact of each variables indicating contributions of the trained model. Fig. 4(e) shows the predicted and real values of BOHB. The result shows some scattering due to small sample size, but verifies correlation between estimated and actual values. Fig. 4(f) shows an E-nose system using commercial chemical gas sensors for non-invasive diagnosis of diseases including diabetes [28]. Saidi et al collected breath samples from 44 volunteers and analyzed them with machine learning algorithms. Fig. 4(g) shows the schematic of 22 VOCs of breath samples, including subjects with chronic kidney disease (CKD) and diabetes mellitus (DM), healthy subjects with high creatinine (HSHS), and healthy subjects with low creatinine (HSLS). The dataset was distinguished by PCA. Fig. 4(h) shows PCA plot of breath VOC samples collected from the E-nose system and other samples for model verification. The results show that E-nose system can identify diseases without overlap using the PCA model.

Fig. 4. 
E-nose systems for diabetes diagnosis. (a) Schematic of the breath analyzer, (b) operation process, and (c) strategy using RNN for BOHB estimation. (d) Feature coefficients of each of variables contributing to regression output. (e) Comparison between predicted BOHB values of the model and truth values. Reprinted with permission from Ref [27]. Copyrights (2022) Elsevier. (f) E-nose measurement setup for diagnosis of CKD, DM, and HS. (g) VOC patterns detected in exhaled breath of CKD, DM, HSHC, and HSLC. (h) Three-dimensional PCA plot of PCA model and other measurement set. Reprinted with permission from Ref [28]. Copyrights (2018) Elsevier.

Bahos et al. fabricated zeolitic imidazolate frameworks (ZIFs) nanocrystal-based surface acoustic wave (SAW) E-nose system [29], which includes four SAW sensors; ZIF-8, ZIF-67, ZIF-8/AuNP, and ZIF-67/AuNP. SAW/ZIF E-nose exhibited high sensitivity, selectivity, and reproducibility to acetone, ethanol, and ammonia.

Gupta et al. presented a simulation study of breath analysis based on MEMS cantilever sensor array [30]. The MEMS cantilever sensors are bent by vapor sorption in the volatile-selective polymer and detect a change in the cantilever resonance frequency. By adopting data mining methods, they successfully detected acetone, isoprene, and ethane with various compounds in virtual breaths.

Chen et al. fabricated a flexible E-nose by synthesizing graphene oxide (GO) using different types of metal ions (M^x+) with different ratio [31]. Fig. 5(a) shows SEM images of reduced graphene oxide-metal (rGO-M) hybrid sensing materials. The rGO-M formed rough, undulating, and porous films, which enhanced sensing performances. The E-nose system including eight rGO-M sensors was tested with exhaled breath collected from 108 volunteers. Fig. 5(b) shows Langmuir-Hill model of the responses toward acetone, isoprene, ammonia, and hydrothion at sub-ppm levels. Using PCA, the rGO-M arrays successfully discriminated four biomarkers as shown in Fig. 5(c). Fig. 5(d) shows result of clinical EB analysis by linear discrimination analysis (LDA). The control group was accurately distinguished from lung cancer patients. Li et al. designed E-nose system using four types of 14 gas sensors for lung cancer diagnosis [32]. They collected gas pattern data using 52 breath samples in various diseased and healthy volunteers. The E-nose system performed classification using LDA with fuzzy 5-NN classifiers. The sensitivity, specificity, and accuracy of discrimination were 91.58, 91.72, and 91.59%, respectively. Kononov et al. fabricated E-nose system consisting of six metal oxide gas sensors [33]. They operated the E-nose system with three different temperature regimes, which can generate different responses patterns. In this study, 118 breath samples from volunteers were used and analyzed using various models, such as k-nearest neighbors (kNN), logistic regression, random forest, LDA, and support vector machine (SVM). The sensitivity, specificity, and accuracy of the classification were 95.0, 100, and 97.2, respectively.

Fig. 5. 
E-nose systems for lung cancer diagnosis. (a) SEM images of GO using different types of metal ions with different ratio of GO to M^x+. (b) Langmuir-Hill fitting curves of rGO-M array for acetone, isoprene, NH₃, and H₂S, respectively. (c) PCA plots of rGO-M array of normalized responses to target gases. (d) LDA result and dynamic responses of EB samples. Reprinted with permission from Ref [31]. Copyrights (2020) American Chemical Society.

Zhong et al constructed a colorimetric sensor array using diverse chemo-responsive dyes [34]. Thirty-six colorimetric sensors generated 108 dimensional vectors with RGB color difference values, which represented responses to different VOCs. The arrays successfully discriminated between 20 VOCs with an accuracy of at least 90%.

Jang et al showed the E-nose that could be a platform for early screening and prognosis monitoring of lung cancer patients through performing the clinical test, according to a procedure delicately prepared by thoracic surgeons [35]. The E-nose system successfully distinguished lung cancer patients from healthy people with 75.0% accuracy by using a multi-layer perceptron (MLP).