# Fileset

[sensors-21-04742.pdf](https://mdr.nims.go.jp/filesets/d97b8d51-51fc-44bd-ab8b-2d93457cbf71/download)

## Creator

Katsushige Inada, Hiroshi Kojima, Yukiko Cho-Isoda, [Ryo Tamura](https://orcid.org/0000-0002-0349-358X), [Gaku Imamura](https://orcid.org/0000-0002-3130-7190), [Kosuke Minami](https://orcid.org/0000-0003-4145-1118), Takahiro Nemoto, [Genki Yoshikawa](https://orcid.org/0000-0002-9136-8964)

## Rights

[Creative Commons BY Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/)

## Other metadata

[Statistical Evaluation of Total Expiratory Breath Samples Collected throughout a Year: Reproducibility and Applicability toward Olfactory Sensor-Based Breath Diagnostics](https://mdr.nims.go.jp/datasets/882ef4bf-a577-454f-9eea-f0858de5189c)

## Fulltext

Statistical Evaluation of Total Expiratory Breath Samples Collected throughout a Year: Reproducibility and Applicability toward Olfactory Sensor-Based Breath DiagnosticssensorsArticleStatistical Evaluation of Total Expiratory Breath SamplesCollected throughout a Year: Reproducibility and Applicabilitytoward Olfactory Sensor-Based Breath DiagnosticsKatsushige Inada 1, Hiroshi Kojima 1,2, Yukiko Cho-Isoda 1, Ryo Tamura 3,4,5, Gaku Imamura 3 ,Kosuke Minami 6 , Takahiro Nemoto 7 and Genki Yoshikawa 7,8,*�����������������Citation: Inada, K.; Kojima, H.;Cho-Isoda, Y.; Tamura, R.; Imamura,G.; Minami, K.; Nemoto, T.;Yoshikawa, G. Statistical Evaluationof Total Expiratory Breath SamplesCollected throughout a Year:Reproducibility and Applicabilitytoward Olfactory Sensor-BasedBreath Diagnostics. Sensors 2021, 21,4742. https://doi.org/10.3390/s21144742Academic Editor: James F. RuslingReceived: 28 April 2021Accepted: 7 July 2021Published: 11 July 2021Publisher’s Note: MDPI stays neutralwith regard to jurisdictional claims inpublished maps and institutional affil-iations.Copyright: © 2021 by the authors.Licensee MDPI, Basel, Switzerland.This article is an open access articledistributed under the terms andconditions of the Creative CommonsAttribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).1 Department of Medical Oncology, Ibaraki Prefectural Central Hospital, Kasama 309-1793, Ibaraki, Japan;k-inada@chubyoin.pref.ibaraki.jp (K.I.); h-kojima@chubyoin.pref.ibaraki.jp (H.K.);y-chou@chubyoin.pref.ibaraki.jp (Y.C.-I.)2 Ibaraki Clinical Education and Training Center, University of Tsukuba Hospital, Kasama 309-1793,Ibaraki, Japan3 World Premier International (WPI) Research Center for Materials Nanoarchitectonics (MANA), NationalInstitute for Materials Science (NIMS), Tsukuba 305-0044, Ibaraki, Japan; TAMURA.Ryo@nims.go.jp (R.T.);IMAMURA.Gaku@nims.go.jp (G.I.)4 Graduate School of Frontier Sciences, The University of Tokyo, Chiba 277-8568, Japan5 Research and Services Division of Materials Data and Integrated System (MaDIS), National Institute forMaterials Science (NIMS), Tsukuba 305-0044, Japan6 International Center for Young Scientists (ICYS), National Institute for Materials Science (NIMS),Tsukuba 305-0044, Ibaraki, Japan; MINAMI.Kosuke@nims.go.jp7 Center for Functional Sensor & Actuator (CFSN), Research Center for Functional Materials, National Institutefor Materials Science (NIMS), Tsukuba 305-0044, Japan; NEMOTO.Takahiro@nims.go.jp8 Materials Science and Engineering, Graduate School of Pure and Applied Science, University of Tsukuba,Tsukuba 305-8571, Ibaraki, Japan* Correspondence: YOSHIKAWA.Genki@nims.go.jp; Tel.: +81-29-860-4749Abstract: The endogenous volatile organic compounds (VOCs) in exhaled breath can be promisingbiomarkers for various diseases including cancers. An olfactory sensor has a possibility for extractinga specific feature from collective variations of the related VOCs with a certain health condition.For this approach, it is important to establish a feasible protocol for sampling exhaled breath inpractical conditions to provide reproducible signal features. Here we report a robust protocol for thebreath analysis, focusing on total expiratory breath measured by a Membrane-type Surface stressSensor (MSS), which possesses practical characteristics for artificial olfactory systems. To assessits reproducibility, 83 exhaled breath samples were collected from one subject throughout morethan a year. It has been confirmed that the reduction of humidity effects on the sensing signalseither by controlling the humidity of purging room air or by normalizing the signal intensitiesleads to reasonable reproducibility verified by statistical analyses. We have also demonstrated theapplicability of the protocol for detecting a target material by discriminating exhaled breaths collectedfrom different subjects with pre- and post-alcohol ingestion on different occasions. This simple yetreproducible protocol based on the total expiratory breath measured by the MSS olfactory sensorswill contribute to exploring the possibilities of clinical applications of breath diagnostics.Keywords: Membrane-type Surface stress Sensor (MSS); artificial olfactory sensor; nanomechanicalsensor; breath analysis; reproducibility; humidity1. IntroductionBreath analysis has long been recognized as an ideal non-invasive diagnostic tech-nique, which poses a potential for its future usage in disease detection and therapeuticmonitoring [1]. Besides nitrogen, oxygen, carbon dioxide, and water vapor, human breathcontains various kinds of volatile organic compounds (VOCs) [2] typically at the ppm toSensors 2021, 21, 4742. https://doi.org/10.3390/s21144742 https://www.mdpi.com/journal/sensorshttps://www.mdpi.com/journal/sensorshttps://www.mdpi.comhttps://orcid.org/0000-0002-3130-7190https://orcid.org/0000-0003-4145-1118https://orcid.org/0000-0002-9136-8964https://doi.org/10.3390/s21144742https://doi.org/10.3390/s21144742https://creativecommons.org/https://creativecommons.org/licenses/by/4.0/https://creativecommons.org/licenses/by/4.0/https://doi.org/10.3390/s21144742https://www.mdpi.com/journal/sensorshttps://www.mdpi.com/article/10.3390/s21144742?type=check_update&version=1Sensors 2021, 21, 4742 2 of 15ppb range, while others may be even in ppt or lower concentration ranges [3]. Exhaledbreath composition reflects the volatile compounds that are emitted from the membraneof the cells and/or from the surrounding microenvironment to the blood, although somevolatile compounds originate in the airways, not being present in the blood [4]. Since someparts of VOCs in exhaled breath are produced endogenously through physiological orpathological processes, the composition of VOCs is believed to reflect a comprehensivemetabolic state of the body. Indeed, several lines of evidence have revealed that certaintypes of VOCs could be related to various diseases [5]. It is thus rational to consider thatpatterns of VOCs in exhaled breath could be a predictive biomarker for disease detection.This notion has prompted many researchers to develop sensitive and reliable measurementtechniques which are applicable to disease detection, especially early cancer detection [6].Numerous studies have thus far reported that some exhaled VOCs are associated withcancer in terms of exhalation kinetics of VOCs [7], sensing response patterns of artificiallyintelligent nanoarrays [8], and diagnostic accuracy of breath tests [9], as well as specificcorrelations such as 15 VOCs for colorectal cancer [10], 3 VOCs for head and neck can-cer [11], 12 VOCs for esophageal and gastric adenocarcinoma [12], 14 VOCs for breastcancer [13], and 22 VOCs for hepatocellular carcinoma [14]. However, clinical practice formeasuring VOCs in exhaled breath has not been established yet. One of the major obstaclesis that the majority of these approaches require a high level of expertise and expensiveequipment such as gas chromatography-mass spectrometry (GC-MS) or various real-timeanalysis techniques including selected ion flow tube mass spectrometry and proton transferreaction-mass spectrometry [15]. A simple and inexpensive artificial olfactory system,which detects multi-dimensional signals generated by multiple sensors array and analysesthem by means of machine learning algorithm, has thus attracted much interest becauseof its applicability to on-site clinical applications as well as its ability to recognize thecomplicated patterns of VOCs as a whole.A Membrane-type Surface stress Sensor (MSS) is a piezoresistive nanomechanicalsensor which possesses potential as a sensing platform in an artificial olfactory system.The MSS consists of an adsorbate membrane suspended by four sensing beams, on whichpiezoresistors are embedded [16]. Response signals transduced by mechanical stress/strain,which is induced by sorption of gas molecules in the coating films coated on the membrane,exert more than 100-fold higher sensitivity in comparison with a standard piezoresistivecantilever-type sensor [17]. Recent studies have demonstrated the potential ability of MSSas an olfactory system in some different analysis, in settings such as quantification ofalcohol contents from vapours of various types of liquids [18], quantification of a ternarymixture consisting of water, ethanol, and methanol [19], discrimination between spicesand herbs [20] even with free-hand measurements [21], and detection of small amountsof VOCs contained in body odour [22]. To our interest, a pilot study analysing exhaledbreath using MSS showed the possibility that patients with head and neck cancer can bediscriminated from healthy volunteers [23] and patients before surgery from those aftersurgery as well as healthy control persons [24]. Although these previous studies havedemonstrated promising sensitivity and specificity of MSS, it is known that the conditionsof samples (e.g., temperature, humidity, and interfering gases) affect the signals measuredby artificial olfactory systems based on chemical sensor arrays, including MSS. Thus, toassess the reproducibility and applicability of MSS olfactory sensors to the practical breath,a statistical evaluation of a large number of breath samples collected for a prolonged periodis required.In this study, we have conducted a statistical evaluation of breath samples collectedthroughout more than a year with controlled measurement conditions. For the breathsampling, we utilized so-called “total expiratory breath sampling” (also known as “mixedexpiratory breath sampling”) [25], which is a simple method that can be practically im-plemented, thereby having a high potential to be standardized. In this method, the totalexhaled breath is collected including dead-space air which is not involved in gaseousexchange in a lung. Although the other two major methods—“late expiratory breathSensors 2021, 21, 4742 3 of 15sampling” and “end-tidal breath sampling”—can reduce the concentrations of exogenousgases by discarding the dead-space air, there is still no known optimal exclusion time orvolume for breath analyses and diagnostics. In addition to the difficulties in controllingsuch conditions for each individual having distinct physiological properties, these meth-ods may miss the airway gases that may contain an important breath marker, such asnitric oxide that has been adopted as the diagnosis and treatment of asthma [26]. Basedon previous reports [4,25], the characteristics of these three methods are summarized inTable 1. The total expiratory breath contains all potential endogenous gases and there area limited number of control parameters for the breath sampling, such as breath holdingtime, which can be rather easily controlled. Therefore, in this study, we focus on thissimple approach and evaluate its reproducibility by analysing the sensing data collectedthroughout more than a year and the applicability to the detection of a testing material inexhaled breath using a same MSS standard measurement module. We have confirmed thehigh reproducibility of the present protocol achieved through the reduction of the majordisturbances (i.e., interfering gases and humidity) in the total expiratory breath samplingby controlling measurement conditions and utilizing a simple data analysis method. Asfor the applicability of this breath measurement system, we demonstrated the detection oftesting material, ethanol, in exhaled breath. Because of the simplicity and the possibilityfor standardization as well as the demonstrated reproducibility and applicability, the pro-posed artificial olfactory system based on the total expiratory breath measured by the MSSmeasurement module can be a promising candidate for future breath diagnostics.Table 1. Characteristics of the three methods for collecting exhaled breath [4,25].Total Expiratory Breath(Mixed Expiratory Breath)Applied in this StudyLate Expiratory Breath End-Tidal Breath(Alveolar Breath)Standardized procedure None None NoneOperation Simple Relatively complicated Relatively complicatedCollecting exhaledbreath phasesAll phases(Dead space + Transition +Alveolar)Partial phases(Minimum dead space +Transition + Alveolar)Partial phase(Alveolar)Discarding exhalation time None First few seconds(Estimated dead space)Initial portion(Monitoring CO2 level)Interfering exogenousvolatile organic compounds Large Moderate Minimal2. Materials and Methods2.1. Study DesignFor verifying the reproducibility of the proposed breath measurement system withvarious disturbances including the long-term drift of the sensing device and the seasonalvariations of the ambient air, exhaled breath samples were repetitively collected in in-door environments from one male volunteer (age 34 to 36) on different days betweenJuly 2018 and December 2019 and all measured by the same MSS measurement module.For examining whether the MSS measurement module can detect a certain component inexhaled breath, ethanol was used as a testing material. Three male volunteers (age 34 to 42,mean 36.7) including one smoker, participated in the alcohol ingestion experiment betweenDecember 2018 and August 2019. Figure 1 shows the flow chart with the timeline in the ex-periments. This study was approved by the institutional review board of Ibaraki PrefecturalCentral Hospital, and informed consent was obtained from all individual participants.Sensors 2021, 21, 4742 4 of 15Sensors 2021, 21, x FOR PEER REVIEW 4 of 16   ly 2018 and December 2019 and all measured by the same MSS measurement module. For examining whether the MSS measurement module can detect a certain component in exhaled breath, ethanol was used as a testing material. Three male volunteers (age 34 to 42, mean 36.7) including one smoker, participated in the alcohol ingestion experiment between December 2018 and August 2019. Figure 1 shows the flow chart with the time-line in the experiments. This study was approved by the institutional review board of Ibaraki Prefectural Central Hospital, and informed consent was obtained from all indi-vidual participants.  Figure 1. Flow chart of the exhaled breath assay with an outline of the protocol along the timeline. MSS stands for Membrane-type Surface stress Sensor. 2.2. MSS Receptor Layers and Measurement Module The MSS standard measurement module produced through the industry-academia-government framework called “MSS Alliance” and “MSS Forum” was used [27]. The working principle of MSS as well as its schematic illustration and an optical microscope image of an MSS chip are presented in Figure 2. The detailed fabrication process of an MSS chip is explained in previous reports [16,17]. The MSS chips used in the present study were purchased from NanoWorld AG. An inkjet spotter (LaboJet-500SP, MICRO-JET Corporation) equipped with a nozzle (IJHBS-300, MICROJET Corporation) was used to coat each receptor material. In this study, poly(4-methylstyrene), poly(2,6-diphenyl-1,4-phenylene oxide), and polymethyl methacrylate were used as receptor materials of Ch 1, 2, and 3, respectively. Each polymer was dissolved in N,N-dimethylformamide (DMF; 1 mg/mL), and the resulting solutions were deposited onto each channel of the MSS. Figure 1. Flow chart of the exhaled breath assay with an outline of the protocol along the timeline. MSS stands forMembrane-type Surface stress Sensor.2.2. MSS Receptor Layers and Measurement ModuleThe MSS standard measurement module produced through the industry-academia-government framework called “MSS Alliance” and “MSS Forum” was used [27]. Theworking principle of MSS as well as its schematic illustration and an optical microscopeimage of an MSS chip are presented in Figure 2. The detailed fabrication process of anMSS chip is explained in previous reports [16,17]. The MSS chips used in the presentstudy were purchased from NanoWorld AG. An inkjet spotter (LaboJet-500SP, MICROJETCorporation) equipped with a nozzle (IJHBS-300, MICROJET Corporation) was used tocoat each receptor material. In this study, poly(4-methylstyrene), poly(2,6-diphenyl-1,4-phenylene oxide), and polymethyl methacrylate were used as receptor materials of Ch1, 2, and 3, respectively. Each polymer was dissolved in N,N-dimethylformamide (DMF;1 mg/mL), and the resulting solutions were deposited onto each channel of the MSS.Sensors 2021, 21, 4742 5 of 15Sensors 2021, 21, x FOR PEER REVIEW 5 of 16   The MSS chips coated with the receptor layers were placed in a Teflon chamber, and the chamber was carefully sealed with O-rings. The chamber was connected to a gas flow system consisting of a switching valve connected with sampling and purging gas lines. The sample and purge gas flows were controlled by an aspiration pump with a flow rate adjusted to 30 mL/min. Data were measured at the bridge voltage of –1.0 V, and the relative resistance changes of piezoresistors were acquired at a sampling fre-quency of 100 Hz. Temperature and relative humidity (RH) of the sample and purge gases are monitored by a temperature/humidity sensor installed in the MSS measure-ment module. In the present study, the same module has been used for all the measure-ments without changing any components.   Figure 2. (a) Optical microscope image of an MSS chip. (b) Schematic illustration of an MSS chip with electrical connec-tions. R indicates a piezoresistor, and VB, Vout, and GND represent the connections to bridge voltage, output signal, and ground, respectively. (c) Working principle of MSS. 2.3. Sample Collection Each exhaled breath sample was collected into a new 1000 mL polymer film sam-pling bag with a valve sleeve (Smart Bag PA; GL Sciences, Tokyo, Japan) using the fol-lowing protocol. After 10 min of rest in the sampling room, the subject’s mouth was rinsed twice with 20 mL of normal saline. The subject inhaled and exhaled for 3 s each and repeated this cycle for three times (in total (3 + 3) × 3 = 18 [s] breathing) through the mouth, and inhaled for 2 s, and held the breath for 6 s, and then, slowly exhaled approx-imately 800 mL of the breath into the sampling bag through a polytetrafluoroethylene (PTFE) tube (50 mm length, 6 mm outer diameter and 3 mm inner diameter) (Flon Indus-try, Tokyo, Japan) via a connector (APU6; Nihon Pisco, Nagano, Japan). Along with breath sampling, room air in the breath sampling room was collected directly into an-other sampling bag by indirect negative pressure sampling to use as a purge gas to re-duce the effects of the interfering gases. The obtained gas samples were stored in an in-cubator (SLC-25A; Mitsubishi Electric Engineering, Tokyo, Japan) at 25 °C. Each bag was disposed of after the measurement and was not reused. Figure 2. (a) Optical microscope image of an MSS chip. (b) Schematic illustration of an MSS chip with electrical connections.R indicates a piezoresistor, and VB, Vout, and GND represent the connections to bridge voltage, output signal, and ground,respectively. (c) Working principle of MSS.The MSS chips coated with the receptor layers were placed in a Teflon chamber, andthe chamber was carefully sealed with O-rings. The chamber was connected to a gasflow system consisting of a switching valve connected with sampling and purging gaslines. The sample and purge gas flows were controlled by an aspiration pump with a flowrate adjusted to 30 mL/min. Data were measured at the bridge voltage of −1.0 V, andthe relative resistance changes of piezoresistors were acquired at a sampling frequencyof 100 Hz. Temperature and relative humidity (RH) of the sample and purge gases aremonitored by a temperature/humidity sensor installed in the MSS measurement module.In the present study, the same module has been used for all the measurements withoutchanging any components.2.3. Sample CollectionEach exhaled breath sample was collected into a new 1000 mL polymer film samplingbag with a valve sleeve (Smart Bag PA; GL Sciences, Tokyo, Japan) using the followingprotocol. After 10 min of rest in the sampling room, the subject’s mouth was rinsed twicewith 20 mL of normal saline. The subject inhaled and exhaled for 3 s each and repeatedthis cycle for three times (in total (3 + 3) × 3 = 18 [s] breathing) through the mouth, andinhaled for 2 s, and held the breath for 6 s, and then, slowly exhaled approximately 800 mLof the breath into the sampling bag through a polytetrafluoroethylene (PTFE) tube (50 mmlength, 6 mm outer diameter and 3 mm inner diameter) (Flon Industry, Tokyo, Japan) viaa connector (APU6; Nihon Pisco, Nagano, Japan). Along with breath sampling, room airin the breath sampling room was collected directly into another sampling bag by indirectnegative pressure sampling to use as a purge gas to reduce the effects of the interferinggases. The obtained gas samples were stored in an incubator (SLC-25A; Mitsubishi ElectricEngineering, Tokyo, Japan) at 25 ◦C. Each bag was disposed of after the measurement andwas not reused.Sensors 2021, 21, 4742 6 of 152.4. Assay by the MSS Measurement ModuleAll the measurements by the MSS measurement module were performed in an in-cubator kept at 25 ◦C. Immediately before the assay, the MSS measurement module waswarmed up to 29–30 ◦C while flowing pure nitrogen gas (over 99.999% purity) through thesensors and gas flow lines. The sampling bags containing exhaled breath and the room airwere stored at 25 ◦C for 30 min after the sampling. The sampling bags were connected tothe MSS measurement module through the sample- and purge-injection lines, respectively(Figure 3a,b). To observe the dependence of the cycle time on the signals, the injectionsequence was set with various cycle times as follows: 40 cycles of 5 s, 10 cycles of 10 s, andone cycle of 30 s exhaled breath sampling and room air purging (Figure 4a). Temperatureand RH of the sample and purge gases were determined at the final cycle of the sequenceusing a temperature/humidity sensor in the MSS measurement module. For the exhaledbreath samples in this study, the measured temperature and RH were almost constantwithin the ranges of 29.4–30.2 ◦C and 70.3–75.3%, respectively.Sensors 2021, 21, x FOR PEER REVIEW 6 of 16   2.4. Assay by the MSS Measurement Module All the measurements by the MSS measurement module were performed in an in-cubator kept at 25 °C. Immediately before the assay, the MSS measurement module was warmed up to 29–30 °C while flowing pure nitrogen gas (over 99.999% purity) through the sensors and gas flow lines. The sampling bags containing exhaled breath and the room air were stored at 25 °C for 30 min after the sampling. The sampling bags were connected to the MSS measurement module through the sample- and purge-injection lines, respectively (Figure 3a,b). To observe the dependence of the cycle time on the sig-nals, the injection sequence was set with various cycle times as follows: 40 cycles of 5 s, 10 cycles of 10 s, and one cycle of 30 s exhaled breath sampling and room air purging (Figure 4a). Temperature and RH of the sample and purge gases were determined at the final cycle of the sequence using a temperature/humidity sensor in the MSS measure-ment module. For the exhaled breath samples in this study, the measured temperature and RH were almost constant within the ranges of 29.4–30.2 °C and 70.3–75.3%, respec-tively.  Figure 3. (a) Measurement setup for gas samples with the MSS measurement module. The module and gas samples were placed in an incubator set at 25 °C. (b) Inlet and outlet of gases on the mod-ule. Figure 3. (a) Measurement setup for gas samples with the MSS measurement module. The moduleand gas samples were placed in an incubator set at 25 ◦C. (b) Inlet and outlet of gases on the module.Sensors 2021, 21, x FOR PEER REVIEW 7 of 16    Figure 4. An example of an output waveform measured by the MSS measurement module. (a) A sequence of output waveforms. The last five of the 10 s cycles (denoted as T1–T5) were selected for the analysis. The black and white arrows indicate the measurement points of temperature and relative humidity (RH), in the sample and purge gases, respective-ly. (b,c) Examples of the output waveforms processed by the start-point offset method (b) and by the min-max normali-zation method (c). Data extraction areas were set as indicated by A–K. 2.5. Verification of Reproducibility Exhaled breath samples collected repetitively from one subject after at least 90 min fasting were analysed by the MSS measurement module. Headspace gas in an airtight vial containing sterile water was used as a reference, namely the humidified gas sample without interfering VOCs. In detail, nitrogen gas was bubbled into 10 mL of water through a PTFE tube (1/16-inch outer diameter, 1 mm inner diameter) which was con-nected to an aluminium bag (GL Sciences, Tokyo, Japan) pre-filled with nitrogen gas, and a headspace tube was directly connected to the sample injection line of the MSS measurement module. For the reference water headspace gas, it was confirmed that the measured temperature and RH in the MSS measurement module were almost constant within the ranges of 29.6–30.4 °C and 70.8–75.1%, respectively. 2.6. Detection of a Testing Material After at least 90 min fasting, subjects took a dose of whiskey (40% v/v) equivalent to 0.5 g/kg body weight of ethanol in 30 min. Breath samples were collected at 0 and 60 min after the alcohol ingestion. During this experiment, subjects were allowed to take only water. Exhaled breath samples were analysed by both the MSS measurement module and the Kitagawa-type ethanol detector tube (104SB with AP-20 aspirating pump; Komyo Rikagaku Kogyo, Kanagawa, Japan), which detects 20–300 ppm of ethanol through the chemical reaction. This experiment was performed for three subjects and each subject participated on three different days. 2.7. Statistical Analysis For analysing the responding signals measured by the MSS measurement module, the last five waveforms in the 10 s cycle area were used (Figure 4a). The waveforms were processed by both the start-point offset method and the min-max normalization method Figure 4. An example of an output waveform measured by the MSS measurement module. (a) A sequence of outputwaveforms. The last five of the 10 s cycles (denoted as T1–T5) were selected for the analysis. The black and white arrowsindicate the measurement points of temperature and relative humidity (RH), in the sample and purge gases, respectively.(b,c) Examples of the output waveforms processed by the start-point offset method (b) and by the min-max normalizationmethod (c). Data extraction areas were set as indicated by A–K.Sensors 2021, 21, 4742 7 of 152.5. Verification of ReproducibilityExhaled breath samples collected repetitively from one subject after at least 90 minfasting were analysed by the MSS measurement module. Headspace gas in an airtightvial containing sterile water was used as a reference, namely the humidified gas samplewithout interfering VOCs. In detail, nitrogen gas was bubbled into 10 mL of water througha PTFE tube (1/16-inch outer diameter, 1 mm inner diameter) which was connected to analuminium bag (GL Sciences, Tokyo, Japan) pre-filled with nitrogen gas, and a headspacetube was directly connected to the sample injection line of the MSS measurement module.For the reference water headspace gas, it was confirmed that the measured temperatureand RH in the MSS measurement module were almost constant within the ranges of29.6–30.4 ◦C and 70.8–75.1%, respectively.2.6. Detection of a Testing MaterialAfter at least 90 min fasting, subjects took a dose of whiskey (40% v/v) equivalentto 0.5 g/kg body weight of ethanol in 30 min. Breath samples were collected at 0 and60 min after the alcohol ingestion. During this experiment, subjects were allowed to takeonly water. Exhaled breath samples were analysed by both the MSS measurement moduleand the Kitagawa-type ethanol detector tube (104SB with AP-20 aspirating pump; KomyoRikagaku Kogyo, Kanagawa, Japan), which detects 20–300 ppm of ethanol through thechemical reaction. This experiment was performed for three subjects and each subjectparticipated on three different days.2.7. Statistical AnalysisFor analysing the responding signals measured by the MSS measurement module,the last five waveforms in the 10 s cycle area were used (Figure 4a). The waveforms wereprocessed by both the start-point offset method and the min-max normalization methodas shown in Figure 4b,c, respectively. In the start-point offset method, the signal value atthe starting point of the sample gas measurement (e.g., at 500 s in Figure 5) is subtractedfrom the waveform data. The min-max normalization approach is processed to convert theresponse value of one waveform into a specific range. The signal data of a waveform werenormalized by setting the minimum and maximum points to 0 and 1, respectively. Thecorrected signal data sets were then segmented into a predetermined size of the area with50% overlap (Figure 4b,c, shown by A–K), as the sensors might respond with a differenttime constant to each odorous molecule. Data extraction areas of 1.4 s duration were setwith 0.6 s overlap with the preceding extraction area for each of the waveforms. From eachof the 1.4 s data extraction areas, 8 output signals with 0.2 s intervals were extracted, andthe mean value of these 8 output signals was defined as the feature value of the extractionarea. This analytical process was applied to each of the data extraction areas of the fivewaveforms which were generated by Ch 1–3 of the MSS.To evaluate statistical significance between two sample groups, the Mann-WhitneyU test was applied by using the EZR software version 1.41 (Saitama Medical Center, JichiMedical University, Saitama, Japan). Two-tailed p-values of less than 0.01 were consideredas statistically significant. The reproducibility of the assay was evaluated by the Bland-Altman analysis, which shows the bias (the mean difference between two feature valuesof measurements) and the 95% limits of agreement (bias ± 1.96 standard deviation (SD)).When zero in all extraction areas falls within the 30% confidence interval, namely 1 × SD ofthe bias, the assay was considered to be reproducible. This analysis is a graphical method,which is widely used to assess the reproducibility of continuous measures by plotting thedifference scores of two measurements of the same variable.The results were presented as a box plot or Bland-Altman plot using the GraphPadPrism version 6.03 software (GraphPad Software Inc., San Diego, CA, USA). In this report,we present the results obtained by analysing the T1 waveform, while we confirmed thatthe results were all reproduced in T2–T5 waveforms as well.Sensors 2021, 21, 4742 8 of 15Sensors 2021, 21, x FOR PEER REVIEW 9 of 16    Figure 5. Sensing data processed by the start-point offset method. (a) The waveforms of reference (n = 33) and breath (n = 83) samples. (b) Comparison of feature values between reference and breath samples evaluated by the two-tailed Mann-Whitney U test (***, p < 0.0001). (c) Comparison of feature values among subgroups categorized by RH range of the purge gas: grey for 20–30% RH (reference n = 10, breath n = 24); blue for 30–40% RH (reference n = 12, breath n = 25); and red for 40–50% RH (reference n = 11, breath n = 34). Box plots represent the results of Kruskal-Wallis test and post-hoc Mann-Whitney U test with Bonferroni correction (*, p < 0.01; ***, p < 0.0001); horizontal lines, lower and upper edges of each box, and whiskers corresponds to medians, 25–75 percentile regions, and 5–95 percentile regions, respectively. 3. Results and Discussion 3.1. Overview of the Signals Measured by the Total Expiratory Breath Sampling Method with the MSS Measurement Module Eighty-three exhaled samples and 33 reference water headspace gas samples were measured by the identical MSS measurement module. Figure 5a shows all the wave-forms of exhaled breath and reference samples processed by the start-point offset meth-od for each channel. It is found that the waveforms exhibit similar profiles specific to each sample/channel, whereas there are significant variations in the signal intensity. When these data are expressed by the feature values, the exhaled breath samples show significantly higher values compared to the reference (p < 0.0001) (Figure 5b), indicating that other components included in the breath samples in addition to water were detect-ed. Although we showed only the extraction areas possessing the lowest p-value for each channel in Figure 5b, significant differences of feature values were observed exclu-sively in all extraction areas of Ch 1 and 2 and extraction areas A–F of Ch 3 (data not shown). 3.2. Effect of Humidity Variation in Room Air Used as a Purge Gas on the Sensing Signals A sensing signal in a sample-purge injection cycle reflects the difference between the sample and purge gases, and thus the common gases included in both sample and purge gases have little effect on the sensing signal. Since most of exhaled breath is com-Figure 5. Sensing data processed by the start-point offset method. (a) The waveforms of reference (n = 33) and breath(n = 83) samples. (b) Comparison of feature values between reference and breath samples evaluated by the two-tailedMann-Whitney U test (***, p < 0.0001). (c) Comparison of feature values among subgroups categorized by RH range of thepurge gas: grey for 20–30% RH (reference n = 10, breath n = 24); blue for 30–40% RH (reference n = 12, breath n = 25); andred for 40–50% RH (reference n = 11, breath n = 34). Box plots represent the results of Kruskal-Wallis test and post-hocMann-Whitney U test with Bonferroni correction (*, p < 0.01; ***, p < 0.0001); horizontal lines, lower and upper edges of eachbox, and whiskers corresponds to medians, 25–75 percentile regions, and 5–95 percentile regions, respectively.3. Results and Discussion3.1. Overview of the Signals Measured by the Total Expiratory Breath Sampling Method with theMSS Measurement ModuleEighty-three exhaled samples and 33 reference water headspace gas samples weremeasured by the identical MSS measurement module. Figure 5a shows all the wave-forms of exhaled breath and reference samples processed by the start-point offset methodfor each channel. It is found that the waveforms exhibit similar profiles specific to eachsample/channel, whereas there are significant variations in the signal intensity. Whenthese data are expressed by the feature values, the exhaled breath samples show signif-icantly higher values compared to the reference (p < 0.0001) (Figure 5b), indicating thatother components included in the breath samples in addition to water were detected.Although we showed only the extraction areas possessing the lowest p-value for eachchannel in Figure 5b, significant differences of feature values were observed exclusively inall extraction areas of Ch 1 and 2 and extraction areas A–F of Ch 3 (data not shown).3.2. Effect of Humidity Variation in Room Air Used as a Purge Gas on the Sensing SignalsA sensing signal in a sample-purge injection cycle reflects the difference between thesample and purge gases, and thus the common gases included in both sample and purgegases have little effect on the sensing signal. Since most of exhaled breath is composedof the inhaled room air as well as a small portion of endogenous VOCs, it is effective touse room air as a purge gas to reduce the signals induced by uncertain interfering gasesSensors 2021, 21, 4742 9 of 15in room air, thereby enhancing the contributions of the endogenous VOCs to the sensingsignal. Accordingly, as explained in the previous section, we used room air as a purgegas in the present study. In this case, however, it is difficult to reduce the contribution ofhumidity to the sensing signal because an exhaled breath is humidified when kept in abody (e.g., in an airway and a lung) at a certain range (in this assay, 71.2–75.3% for themeasured 83 breath samples; cf. 70.8–75.1% for the measured 33 reference samples), whilethe humidity in room air can fluctuate in a much wider range (20.1–50.0% for the measured83 and 33 samples).To explicitly confirm the effect of humidity on the sensing signals, the reference andthe exhaled breath samples were categorized into three subgroups according to the RHrange in the purge gas, namely 20–30% (reference n = 10, breath n = 24), 30–40% (referencen = 12, breath n = 25), and 40–50% (reference n = 11, breath n = 34) subgroups. The overalldifference among subgroups was evaluated by the Kruskal-Wallis test, which gave a p-valueof less than 0.0001 for all extraction areas of each channel. Data of the extraction area withthe lowest p-value are presented as representative data for each channel in Figure 5c. TheMann-Whitney U test with Bonferroni correction showed significant pairwise difference(p < 0.0001) among the RH subgroups if compared within the reference group or exhaledbreath group. Significant differences were also observed between reference and exhaledbreath samples if compared between the same RH subgroups (p < 0.0001 for Ch 1 and 2and p < 0.01 for Ch 3). However, if compared irrespective of the RH subgroup (for example,comparison between 20–30% subgroup of reference in Ch 1 and 40–50% subgroup of breath),significant differences were not always observed between reference and breath samples.These results were consistent for all extraction areas of each channel (data not shown),indicating that the effect of RH in the purge gas cannot be ignored when interpreting thedata processed by the start-point offset method.It should be also noted here that the fluctuation of temperature is another majorfactor that possibly affects analytical results of exhaled breath, as reported previously [28].However, the temperature of a sample is easier to control than the humidity by simplyplacing the sampling bags in a temperature-regulated incubator for a certain period (e.g.,30 min in the present study). In this study, the temperatures of the samples were wellcontrolled within a rather small range (29.4–30.1 ◦C), which was confirmed not to induce asignificant effect on the sensing signals.3.3. Evaluation of Reproducibility through Humidity Correction and Signal Intensity NormalizationTo evaluate how much reproducibility can be achieved by reducing the humidityeffect, we utilized the following two approaches to reduce the effect of the humidityinconsistency between the sample and purge gases: (1) analysis of paired samples withclose RH conditions and (2) normalization of signal intensity.3.3.1. Analysis of Paired Samples with Close RH ConditionsTo assess whether feature values of exhaled breath processed by the start-point offsetmethod are reproducibly obtained if assayed under close RH conditions, paired samplescollected under almost the same RH conditions (difference no more than 0.1% RH) wereanalysed by the Bland-Altman plots. Considering that breath samples, although collectedfrom one same volunteer, are not identical because of biological variations of the body andeffects of ingested foods, an allowable limit of the bias was set at 30% confidence intervals.Figure 6 depicts the Bland-Altman plots of all 29 paired samples for the extraction areasA, F, and K of Ch 1–3 as representative data. The analysis covering all extraction areas,namely areas A–K, showed that 89.7–93.1% of pairs for Ch 1 and 3 and 96.6% of pairs forCh 2 are plotted within the 95% limits of agreement. The bias of the assay was from –0.0250to –0.0025 for Ch 1, from –0.0228 to –0.0104 for Ch 2, and from 0.0831 to 0.1079 for Ch 3.Furthermore, the 30% confidence intervals always included the zero for all extraction areasof Ch 1–3, indicating that the measurements of the exhaled breath by the MSS measurementSensors 2021, 21, 4742 10 of 15module exert high reproducibility when the humidity of room air used as a purge gas iswell controlled at a certain value that is consistent in each assay.Sensors 2021, 21, x FOR PEER REVIEW 11 of 16   1 and 3 and 96.6% of pairs for Ch 2 are plotted within the 95% limits of agreement. The bias of the assay was from –0.0250 to –0.0025 for Ch 1, from –0.0228 to –0.0104 for Ch 2, and from 0.0831 to 0.1079 for Ch 3. Furthermore, the 30% confidence intervals always in-cluded the zero for all extraction areas of Ch 1–3, indicating that the measurements of the exhaled breath by the MSS measurement module exert high reproducibility when the humidity of room air used as a purge gas is well controlled at a certain value that is consistent in each assay.  Figure 6. Difference of feature values between paired samples with the difference no more than 0.1% RH. The differ-ences of feature values between 29 paired samples were evaluated by the Bland-Altman plots. Solid lines indicate the bi-as, and dashed lines correspond to 95% limits of agreement. Shaded areas represent 30% confidence intervals. The re-sults of extraction areas A, F, and K are shown as representative plots. 3.3.2. Normalization of Signal Intensity It is usually challenging to control the humidity of room air in practical conditions. Thus, we also examined the possibility of reducing the humidity effect through the pre-treatment of the measured sensing data. For this purpose, we utilized the min-max nor-malization method. As this method compresses the waveform in between 0 and 1, the information on the signal intensity is mostly eliminated from the sensing data. Since the RH values of the purge gases (i.e., room air) correlate significantly with the signal inten-sity, as confirmed in Figure 5c, the min-max normalization method could reduce the humidity-related effect and relatively enhance the effects of remaining components, es-pecially endogenous VOCs, included only in the breath samples. The Kruskal-Wallis test showed an overall difference among each of the RH sub-groups for all extraction areas of each channel (p < 0.0001). The data of the extraction ar-ea with the lowest p-value by the Kruskal-Wallis test are shown as representative data in Figure 7. Additional pairwise analysis by the Mann-Whitney U test with Bonferroni cor-Figure 6. Difference of feature values between paired samples with the difference no more than 0.1% RH. The differencesof feature values between 29 paired samples were evaluated by the Bland-Altman plots. Solid lines indicate the bias, anddashed lines correspond to 95% limits of agreement. Shaded areas represent 30% confidence intervals. The results ofextraction areas A, F, and K are shown as representative plots.3.3.2. Normalization of Signal IntensityIt is usually challenging to control the humidity of room air in practical conditions.Thus, we also examined the possibility of reducing the humidity effect through the pre-treatment of the measured sensing data. For this purpose, we utilized the min-maxnormalization method. As this method compresses the waveform in between 0 and 1,the information on the signal intensity is mostly eliminated from the sensing data. Sincethe RH values of the purge gases (i.e., room air) correlate significantly with the signalintensity, as confirmed in Figure 5c, the min-max normalization method could reducethe humidity-related effect and relatively enhance the effects of remaining components,especially endogenous VOCs, included only in the breath samples.The Kruskal-Wallis test showed an overall difference among each of the RH subgroupsfor all extraction areas of each channel (p < 0.0001). The data of the extraction area withthe lowest p-value by the Kruskal-Wallis test are shown as representative data in Figure 7.Additional pairwise analysis by the Mann-Whitney U test with Bonferroni correctionshowed that feature values of exhaled breath samples are significantly different from thoseof reference samples (p < 0.01) irrespective of the RH subgroup in extraction area A–J ofCh 1, extraction area B, J, and K of Ch 2, and extraction area A, J, and K of Ch 3. Theseresults suggest that the min-max normalization method can significantly reduce the effectof RH in the purge gas if appropriate extraction areas are selected, providing reproducibleSensors 2021, 21, 4742 11 of 15sensing data with certain information regarding the gaseous components included inbreath samples.Sensors 2021, 21, x FOR PEER REVIEW 12 of 16   rection showed that feature values of exhaled breath samples are significantly different from those of reference samples (p < 0.01) irrespective of the RH subgroup in extraction area A–J of Ch 1, extraction area B, J, and K of Ch 2, and extraction area A, J, and K of Ch 3. These results suggest that the min-max normalization method can significantly reduce the effect of RH in the purge gas if appropriate extraction areas are selected, providing reproducible sensing data with certain information regarding the gaseous components included in breath samples.  Figure 7. Sensing data processed by the min-max normalization method. (a) The waveforms of reference (n = 33) and breath (n = 83) samples. (b) Comparison of feature values between reference and breath samples evaluated by the two-tailed Mann-Whitney U test (***, p < 0.0001). (c) Comparison of feature values among subgroups categorized by RH range of the purge gas: grey for 20–30% RH (reference n = 10, breath n = 24); blue for 30–40% RH (reference n = 12, breath n = 25); and red for 40–50% RH (reference n = 11, breath n = 34). Box plots represent the results of Kruskal-Wallis test and post-hoc Mann-Whitney U test with Bonferroni correction (*, p < 0.01; ***, p < 0.0001). 3.4. Detection of a Testing Material in Exhaled Breath We next examined whether a testing material, ethanol, in exhaled breath is detecta-ble by the MSS measurement module. After one hour of alcohol ingestion, ethanol con-centrations measured by the ethanol detector tube were in the range of 60–120 ppm (Figure 8). It was confirmed that in eighteen breath samples the measured temperature and RH in the MSS measurement module were constant within the ranges of 29.4–30.2 °C and 70.3–73.6%, respectively. For the room air, temperature and RH were observed within the ranges of 29.4–30.2 °C and 23.0–41.8%, respectively. As the humidity levels of room air used as purge gases were also not controlled in these measurements, significant differences in the feature values extracted from the sensing data processed by the start-point offset method were difficult to be observed between pre- and post-alcohol inges-tion samples in all extraction areas of each channel. Thus, we applied the min-max nor-malization method to the obtained data and confirmed significant differences of feature values between pre- and post-alcohol ingestion samples in all extraction areas of Ch 1 Figure 7. Sensing data processed by the min-max normalization method. (a) The waveforms of reference (n = 33) and breath(n = 83) samples. (b) Comparison of feature values between reference and breath samples evaluated by the two-tailedMann-Whitney U test (***, p < 0.0001). (c) Comparison of feature values among subgroups categorized by RH range of thepurge gas: grey for 20–30% RH (reference n = 10, breath n = 24); blue for 30–40% RH (reference n = 12, breath n = 25); andred for 40–50% RH (reference n = 11, breath n = 34). Box plots represent the results of Kruskal-Wallis test and post-hocMann-Whitney U test with Bonferroni correction (*, p < 0.01; ***, p < 0.0001).3.4. Detection of a Testing Material in Exhaled BreathWe next examined whether a testing material, ethanol, in exhaled breath is detectableby the MSS measurement module. After one hour of alcohol ingestion, ethanol concentra-tions measured by the ethanol detector tube were in the range of 60–120 ppm (Figure 8). Itwas confirmed that in eighteen breath samples the measured temperature and RH in theMSS measurement module were constant within the ranges of 29.4–30.2 ◦C and 70.3–73.6%,respectively. For the room air, temperature and RH were observed within the rangesof 29.4–30.2 ◦C and 23.0–41.8%, respectively. As the humidity levels of room air usedas purge gases were also not controlled in these measurements, significant differencesin the feature values extracted from the sensing data processed by the start-point offsetmethod were difficult to be observed between pre- and post-alcohol ingestion samples inall extraction areas of each channel. Thus, we applied the min-max normalization methodto the obtained data and confirmed significant differences of feature values between pre-and post-alcohol ingestion samples in all extraction areas of Ch 1 and in extraction areasD–J of Ch 2. Contrary, a significant difference in the feature values was not observedbetween pre- and post-alcohol ingestion samples in all extraction areas of Ch 3. The ex-traction area with one of the lowest p-value for each channel is shown in Figure 9. Theseresults are consistent with the previous observations that the receptor materials coatedon Ch 1 and 2 are more effective in extracting the information on ethanol vapor becauseSensors 2021, 21, 4742 12 of 15of their hydrophobic characteristics, in contrast to the receptor material on Ch 3 with thehydrophilic property [18]. These analyses suggest that the MSS measurement module withthe appropriate data processing can discriminate testing material contained in exhaledbreath, even if the breath samples were obtained from different examinees on differentoccasions with humidity variations.Sensors 2021, 21, x FOR PEER REVIEW 13 of 16   and in extraction areas D–J of Ch 2. Contrary, a significant difference in the feature val-ues was not observed between pre- and post-alcohol ingestion samples in all extraction areas of Ch 3. The extraction area with one of the lowest p-value for each channel is shown in Figure 9. These results are consistent with the previous observations that the receptor materials coated on Ch 1 and 2 are more effective in extracting the information on ethanol vapor because of their hydrophobic characteristics, in contrast to the receptor material on Ch 3 with the hydrophilic property [18]. These analyses suggest that the MSS measurement module with the appropriate data processing can discriminate test-ing material contained in exhaled breath, even if the breath samples were obtained from different examinees on different occasions with humidity variations.  Figure 8. Breath ethanol concentration of pre- (n = 9) and post- (n = 9) alcohol ingestion. The mean value is represented as a horizontal line. Pre-alcohol ingestion samples exhibited undetectable lev-els (i.e., < 20 ppm).  Figure 9. Feature values of pre- and post-alcohol ingestion breath samples processed by the min-max normalization method. Box plots represent the statistical differences evaluated by the two-tailed Mann-Whitney U test (***, p < 0.0001; ns, not significant). 4. Conclusions In this study, we have established a simple yet robust protocol for breath analysis using the MSS olfactory sensor system. The reproducibility of the proposed breath measurement system based on the total expiratory breath sampling has been confirmed Figure 8. Breath ethanol concentration of pre- (n = 9) and post- (n = 9) alcohol ingestion. The meanvalue is represented as a horizontal line. Pre-alcohol ingestion samples exhibited undetectable levels(i.e., < 20 ppm).Sensors 2021, 21, x FOR PEER REVIEW 13 of 16   and in extraction areas D–J of Ch 2. Contrary, a significant difference in the feature val-ues was not observed between pre- and post-alcohol ingestion samples in all extraction areas of Ch 3. The extraction area with one of the lowest p-value for each channel is shown in Figure 9. These results are consistent with the previous observations that the receptor materials coated on Ch 1 and 2 are more effective in extracting the information on ethanol vapor because of their hydrophobic characteristics, in contrast to the receptor material on Ch 3 with the hydrophilic property [18]. These analyses suggest that the MSS measurement module with the appropriate data processing can discriminate test-ing material contained in exhaled breath, even if the breath samples were obtained from different examinees on different occasions with humidity variations.  Figure 8. Breath ethanol concentration of pre- (n = 9) and post- (n = 9) alcohol ingestion. The mean value is represented as a horizontal line. Pre-alcohol ingestion samples exhibited undetectable lev-els (i.e., < 20 ppm).  Figure 9. Feature values of pre- and post-alcohol ingestion breath samples processed by the min-max normalization method. Box plots represent the statistical differences evaluated by the two-tailed Mann-Whitney U test (***, p < 0.0001; ns, not significant). 4. Conclusions In this study, we have established a simple yet robust protocol for breath analysis using the MSS olfactory sensor system. The reproducibility of the proposed breath measurement system based on the total expiratory breath sampling has been confirmed Figure 9. Feature values of pre- and post-alcohol ingestion breath samples processed by the min-maxnormalization method. Box plots represent the statistical differences evaluated by the two-tailedMann-Whitney U test (***, p < 0.0001; ns, not significant).4. ConclusionsIn this study, we have established a simple yet robust protocol for breath analysisusing the MSS olfactory sensor system. The reproducibility of the proposed breath mea-surement system based on the total expiratory breath sampling has been confirmed by thestatistical evaluation of more than 80 exhaled breath samples collected from one volunteerthroughout more than a year. It has been demonstrated that the key to achieving a reason-able reproducibility is to reduce the undesired effects, such as interfering exogenous gasesand humidity, stemming from the differences between the sample and the purge gases.These two typical inconsistencies between sample and purge gases have been compensatedby adopting the combination of the total expiratory breath sampling and the room airSensors 2021, 21, 4742 13 of 15purge, and by suppressing the contributions of the humidity in a purge gas, respectively.Following the presented protocol, it has been confirmed that a testing material in exhaledbreath is readily detectable. It should be noted that the main limitation of the presentprotocol for practical application is its rather long total assay time, which is about an houras shown in Figure 1. Since the major parts of this assay include the warm-up operation ofthe measurement module and storage of breath sample, the development of an easy-to-usesystem with precise thermal control will significantly reduce the assay time as well as thetotal cost for each analysis.While we have demonstrated the feasibility of assaying exhaled breath samples usingthe MSS olfactory sensor, this is a fundamental model experiment verifying the repro-ducibility and detecting a simple testing material in exhaled breath. Previous studiesmeasuring exhaled breath of cancer patients suggest that diverse kinds of VOCs and theirvariations as a whole with up or down gradients of each component can be a predictivebiomarker. Although the identification of a specific kind of a “marker molecule” willcertainly be a breakthrough in breath diagnosis, the establishment of a “marker feature” ina pattern of multi-dimensional olfactory sensing signals will be another potential way tomake a great leap towards practical implementation. In the latter approach, an easy andreproducible protocol with a feasible measurement device and controllable conditions is theessential requirement because such a “marker feature” must be reproduced by basically anyoperator irrespective of their expertise as well as the measurement conditions including thefluctuations of interfering gases and humidity. Although further investigations are requiredto assess the practical applicability and versatility of the presented assay, we believe thatthis simple protocol together with the MSS olfactory sensor technology can be a promisingoption. Moreover, this approach possesses plenty of room for further improvement in theperformance through the optimization of receptor materials as well as the integration ofadvanced machine learning algorithms.Author Contributions: Conceptualization, K.I., H.K., K.M. and G.Y.; methodology, K.I., Y.C.-I.,K.M., T.N. and G.Y.; software, G.Y.; validation, K.I., H.K., Y.C.-I., R.T., G.I., K.M., T.N. and G.Y.;formal analysis, K.I. and H.K.; investigation, K.I.; resources, H.K. and G.Y.; data curation, K.I.;writing—original draft preparation, K.I. and H.K.; writing—review and editing, R.T., G.I., K.M.and G.Y.; visualization, K.I.; supervision, H.K. and G.Y.; project administration, H.K. and G.Y.;funding acquisition, H.K. and G.Y. All authors have read and agreed to the published version ofthe manuscript.Funding: This research was supported by Japan Society for the Promotion of Science KAKENHI,grant number 18K07318, Grant-in-Aid for Scientific Research (A), 18H04168, Challenging Research(Pioneering), 20K20554, Japanese Ministry of Education, Culture, Sports, Science and Technology,Subsidy based on the three kinds of electric laws, JST CREST (JPMJCR1665), the Center for FunctionalSensor & Actuator (CFSN), NIMS, the World Premier International Research Center Initiative (WPI)on Materials Nanoarchitectonics (MANA), NIMS, and The Public/Private R&D Investment StrategicExpansion Pro-gram (PRISM), Cabinet Office, Japan.Institutional Review Board Statement: The study was conducted according to the guidelines of theDeclaration of Helsinki, and approved by the Institutional Review Board of Institutional ReviewBoard of the Ibaraki Prefectural Central Hospital (protocol code 27-42 and date of approval 26August 2015).Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.Data Availability Statement: The data presented in this study are available on request from thecorresponding author. The data are not publicly available due to privacy restrictions.Acknowledgments: The authors would like to acknowledge the members of “MSS Alliance” and“MSS Forum” for the development of the MSS standard measurement module and for fruitfuldiscussions. The authors thank Yukio Sato, Yusuke Saeki, and Naoki Maki from University ofTsukuba for their helpful discussion, and Toyomi Yokota and Masaaki Matoba from National Institutefor Materials Science for their support in the project.Sensors 2021, 21, 4742 14 of 15Conflicts of Interest: The authors declare no conflict of interest.References1. Miekisch, W.; Schubert, J.K.; Noeldge-Schomburg, G.F. Diagnostic potential of breath analysis—Focus on volatile organiccompounds. Clin. Chim. Acta 2004, 347, 25–39. [CrossRef] [PubMed]2. Hakim, M.; Broza, Y.Y.; Barash, O.; Peled, N.; Phillips, M.; Amann, A.; Haick, H. Volatile Organic Compounds of Lung Cancerand Possible Biochemical Pathways. Chem. Rev. 2012, 112, 5949–5966. [CrossRef]3. Konvalina, G.; Haick, H. Sensors for Breath Testing: From Nanomaterials to Comprehensive Disease Detection. Acc. Chem. Res.2014, 47, 66–76. [CrossRef]4. Pereira, J.; Porto-Figueira, P.; Cavaco, C.; Taunk, K.; Rapole, S.; Dhakne, R.; Nagarajaram, H.; Câmara, J.S. Breath Analysis as aPotential and Non-Invasive Frontier in Disease Diagnosis: An Overview. Metabolites 2015, 5, 3–55. [CrossRef]5. Broza, Y.Y.; Haick, H. Nanomaterial-based sensors for detection of disease by volatile organic compounds. Nanomedicine 2013, 8,785–806. [CrossRef]6. Haick, H.; Broza, Y.Y.; Mochalski, P.; Ruzsanyi, V.; Amann, A. Assessment, origin, and implementation of breath volatile cancermarkers. Chem. Soc. Rev. 2014, 43, 1423–1449. [CrossRef] [PubMed]7. Amann, A.; Mochalski, P.; Ruzsanyi, V.; Broza, Y.Y.; Haick, H. Assessment of the exhalation kinetics of volatile cancer biomarkersbased on their physicochemical properties. J. Breath Res. 2014, 8, 016003. [CrossRef]8. Nakhleh, M.K.; Amal, H.; Jeries, R.; Broza, Y.Y.; Aboud, M.; Gharra, A.; Ivgi, H.; Khatib, S.; Badarneh, S.; Har-Shai, L.; et al.Diagnosis and Classification of 17 Diseases from 1404 Subjects via Pattern Analysis of Exhaled Molecules. ACS Nano 2017, 11,112–125. [CrossRef] [PubMed]9. Hanna, G.B.; Boshier, P.R.; Markar, S.R.; Romano, A. Accuracy and methodologic challenges of volatile organic compound-basedexhaled breath tests for cancer diagnosis: A systematic review and meta-analysis. JAMA Oncol. 2019, 5, e182815. [CrossRef][PubMed]10. Altomare, D.F.; Di Lena, M.; Porcelli, F.; Trizio, L.; Travaglio, E.; Tutino, M.; Dragonieri, S.; Memeo, V.; de Gennaro, G. Exhaledvolatile organic compounds identify patients with colorectal cancer. Br. J. Surg. 2013, 100, 144–150. [CrossRef]11. Gruber, M.L.; Tisch, U.; Jeries, R.; Amal, H.; Hakim, M.; Ronen, O.; Marshak, T.; Zimmerman, D.R.; Israel, O.; Amiga, E.; et al.Analysis of exhaled breath for diagnosing head and neck squamous cell carcinoma: A feasibility study. Br. J. Cancer 2014, 111,790–798. [CrossRef]12. Kumar, S.; Huang, J.; Abbassi-Ghadi, N.; Mackenzie, H.A.; Veselkov, K.A.; Hoare, J.M.; Lovat, L.B.; Španěl, P.; Smith, D.; Hanna,G.B. Mass spectrometric analysis of exhaled breath for the identification of volatile organic compound biomarkers in esophagealand gastric adenocarcinoma. Ann. Surg. 2015, 262, 981–990. [CrossRef] [PubMed]13. Barash, O.; Zhang, W.; Halpern, J.M.; Hua, Q.-L.; Pan, Y.-Y.; Kayal, H.; Khoury, K.; Liu, H.; Davies, M.P.; Haick, H. Differentiationbetween genetic mutations of breast cancer by breath volatolomics. Oncotarget 2015, 6, 44864–44876. [CrossRef] [PubMed]14. Miller-Atkins, G.; Acevedo-Moreno, L.; Grove, D.; Dweik, R.A.; Tonelli, A.R.; Brown, J.M.; Allende, D.S.; Aucejo, F.; Rotroff,D.M. Breath metabolomics provides an accurate and noninvasive approach for screening cirrhosis, primary, and secondary livertumors. Hepatol. Commun. 2020, 4, 1041–1055. [CrossRef] [PubMed]15. Paschke, K.M.; Mashir, A.; Dweik, R.A. Clinical applications of breath testing. F1000 Med. Rep. 2010, 2, 56. [CrossRef] [PubMed]16. Yoshikawa, G.; Akiyama, T.; Gautsch, S.; Vettiger, P.; Rohrer, H. Nanomechanical Membrane-type Surface Stress Sensor. Nano Lett.2011, 11, 1044–1048. [CrossRef]17. Yoshikawa, G.; Akiyama, T.; Loizeau, F.; Shiba, K.; Gautsch, S.; Nakayama, T.; Vettiger, P.; De Rooij, N.F.; Aono, M. TwoDimensional Array of Piezoresistive Nanomechanical Membrane-Type Surface Stress Sensor (MSS) with Improved Sensitivity.Sensors 2012, 12, 15873–15887. [CrossRef] [PubMed]18. Shiba, K.; Tamura, R.; Imamura, G.; Yoshikawa, G. Data-driven nanomechanical sensing: Specific information extraction from acomplex system. Sci. Rep. 2017, 7, 3661. [CrossRef] [PubMed]19. Shiba, K.; Tamura, R.; Sugiyama, T.; Kameyama, Y.; Koda, K.; Sakon, E.; Minami, K.; Ngo, H.T.; Imamura, G.; Tsuda, K.; et al.Functional Nanoparticles-Coated Nanomechanical Sensor Arrays for Machine Learning-Based Quantitative Odor Analysis. ACSSens. 2018, 3, 1592–1600. [CrossRef] [PubMed]20. Imamura, G.; Shiba, K.; Yoshikawa, G. Smell identification of spices using nanomechanical membrane-type surface stress sensors.Jpn. J. Appl. Phys. 2016, 55, 1102B3. [CrossRef]21. Imamura, G.; Shiba, K.; Yoshikawa, G.; Washio, T. Free-hand gas identification based on transfer function ratios without gas flowcontrol. Sci. Rep. 2019, 9, 9768. [CrossRef] [PubMed]22. Minami, K.; Shiba, K.; Yoshikawa, G. Discrimination of structurally similar odorous molecules with various concentrations byusing a nanomechanical sensor. Anal. Methods 2018, 10, 3720–3726. [CrossRef]23. Loizeau, F.; Lang, H.P.; Akiyama, T.; Gautsch, S.; Vettiger, P.; Tonin, A.; Yoshikawa, G.; Gerber, C.; De Rooij, N. Piezoresistivemembrane-type surface stress sensor arranged in arrays for cancer diagnosis through breath analysis. In Proceedings of the2013 IEEE 26th International Conference on Micro Electro Mechanical Systems (MEMS), Taipei, Taiwan, 20–24 January 2013; pp.621–624.http://doi.org/10.1016/j.cccn.2004.04.023http://www.ncbi.nlm.nih.gov/pubmed/15313139http://doi.org/10.1021/cr300174ahttp://doi.org/10.1021/ar400070mhttp://doi.org/10.3390/metabo5010003http://doi.org/10.2217/nnm.13.64http://doi.org/10.1039/C3CS60329Fhttp://www.ncbi.nlm.nih.gov/pubmed/24305596http://doi.org/10.1088/1752-7155/8/1/016003http://doi.org/10.1021/acsnano.6b04930http://www.ncbi.nlm.nih.gov/pubmed/28000444http://doi.org/10.1001/jamaoncol.2018.2815http://www.ncbi.nlm.nih.gov/pubmed/30128487http://doi.org/10.1002/bjs.8942http://doi.org/10.1038/bjc.2014.361http://doi.org/10.1097/SLA.0000000000001101http://www.ncbi.nlm.nih.gov/pubmed/25575255http://doi.org/10.18632/oncotarget.6269http://www.ncbi.nlm.nih.gov/pubmed/26540569http://doi.org/10.1002/hep4.1499http://www.ncbi.nlm.nih.gov/pubmed/32626836http://doi.org/10.3410/M2-56http://www.ncbi.nlm.nih.gov/pubmed/21173863http://doi.org/10.1021/nl103901ahttp://doi.org/10.3390/s121115873http://www.ncbi.nlm.nih.gov/pubmed/23202237http://doi.org/10.1038/s41598-017-03875-7http://www.ncbi.nlm.nih.gov/pubmed/28623343http://doi.org/10.1021/acssensors.8b00450http://www.ncbi.nlm.nih.gov/pubmed/30110149http://doi.org/10.7567/JJAP.55.1102B3http://doi.org/10.1038/s41598-019-46164-1http://www.ncbi.nlm.nih.gov/pubmed/31278339http://doi.org/10.1039/C8AY01224ESensors 2021, 21, 4742 15 of 1524. Lang, H.P.; Loizeau, F.; Hiou-Feige, A.; Rivals, J.-P.; Romero, P.; Akiyama, T.; Gerber, C.; Meyer, E. Piezoresistive MembraneSurface Stress Sensors for Characterization of Breath Samples of Head and Neck Cancer Patients. Sensors 2016, 16, 1149. [CrossRef][PubMed]25. Lawal, O.; Ahmed, W.M.; Nijsen, T.M.E.; Goodacre, R.; Fowler, S.J. Exhaled breath analysis: A review of ‘breath-taking’ methodsfor off-line analysis. Metabolomics 2017, 13, 110. [CrossRef] [PubMed]26. Dweik, R.A.; Boggs, P.B.; Erzurum, S.C.; Irvin, C.G.; Leigh, M.W.; Lundberg, J.O.; Olin, A.-C.; Plummer, A.L.; Taylor, D.R. AnOfficial ATS Clinical Practice Guideline: Interpretation of Exhaled Nitric Oxide Levels (FeNO) for Clinical Applications. Am. J.Respir. Crit. Care Med. 2011, 184, 602–615. [CrossRef] [PubMed]27. MSS Forum. Available online: https://mss-forum.com (accessed on 9 April 2021).28. Mansour, E.; Vishinkin, R.; Rihet, S.; Saliba, W.; Fish, F.; Sarfati, P.; Haick, H. Measurement of temperature and relative humidityin exhaled breath. Sens. Actuators B Chem. 2020, 304, 127371. [CrossRef]http://doi.org/10.3390/s16071149http://www.ncbi.nlm.nih.gov/pubmed/27455276http://doi.org/10.1007/s11306-017-1241-8http://www.ncbi.nlm.nih.gov/pubmed/28867989http://doi.org/10.1164/rccm.9120-11SThttp://www.ncbi.nlm.nih.gov/pubmed/21885636https://mss-forum.comhttp://doi.org/10.1016/j.snb.2019.127371 Introduction  Materials and Methods  Study Design  MSS Receptor Layers and Measurement Module  Sample Collection  Assay by the MSS Measurement Module  Verification of Reproducibility  Detection of a Testing Material  Statistical Analysis  Results and Discussion  Overview of the Signals Measured by the Total Expiratory Breath Sampling Method with the MSS Measurement Module  Effect of Humidity Variation in Room Air Used as a Purge Gas on the Sensing Signals  Evaluation of Reproducibility through Humidity Correction and Signal Intensity Normalization  Analysis of Paired Samples with Close RH Conditions  Normalization of Signal Intensity  Detection of a Testing Material in Exhaled Breath  Conclusions  References