Feasibility of Using the Hexoskin Smart Garment for Natural Environment Observation of Respiration Topography

Background: Limited research has been done to measure ambulatory respiratory behavior, in particular those associated with tobacco use, in the natural environment due to a lack of monitoring techniques. Respiratory topography parameters provide useful information for modeling particle deposition in the lung and assessing exposure risk and health effects associated with tobacco use. Commercially available Wearable Respiratory Monitors (WRM), such as the Hexoskin Smart Garment, have embedded sensors that measure chest motion and may be adapted for measuring ambulatory lung volume. Methods: Self-reported “everyday” and “some days” Hookah and Cigarette smokers were recruited for a 3-day natural environment observation study. Participants wore the Hexoskin shirt while using their preferred tobacco product. The shirt was calibrated on them prior to, during, and after the observation period. A novel method for calculating the calibration parameters is presented. Results: NH = 5 Hookah and NC = 3 Cigarette participants were enrolled. Calibration parameters were obtained and applied to the observed chest motion waveform from each participant to obtain their lung volume waveform. Respiratory topography parameters were derived from the lung volume waveform. Conclusion: The feasibility of using the Hexoskin for measuring ambulatory respiratory topography parameters in the natural environment is demonstrated.


Introduction
Limited research has been conducted to study respiration behavior of tobacco use in the natural environment, in part due to lack of suitable ambulatory monitoring techniques. Current commercially available Wearable Respiratory Monitors (WRM) may be leveraged to measure ambulatory respiration topography in the natural environment 1 although little research 2 has been done to utilize these devices for this application. In this paper we demonstrate feasibility of using of one such WRM, the Hexoskin Smart Garment (Carré Technologies Inc., Montréal, PQ, Canada), to measure ambulatory respiration topography of cigarette and hookah smokers in their natural use environment.
To fully assess the health effects associated with tobacco use, both puffing and respiration behaviors must be considered. Traditionally, researchers have focused only on puffing behavior (puffing topography) quantified by puff flow rate, puff duration, puff volume, and interpuff gap, typically measured using a topography monitor [3][4][5][6] . However, puffing topography by itself is insufficient for understanding the distribution of toxic constituents in the respiratory tract beyond the oral cavity. The flow dynamics of the inhaled and exhaled volume, characterized by the flow rate, volume, duration, start and end time of each inhale and exhale, breath-hold period and time between breaths, are necessary for accurate modelling of deposition of inhaled tobacco particles in the lungs 7 . These parameters are collectively referred to as respiration topography.
Researchers have shown that users may alter their puffing and respiration behavior to compensate for the reduction in nicotine dose, typically by increasing puff and inhalation volume [8][9][10][11] , but other compensatory mechanisms include regulating inhalation flow rate, increasing breath-hold duration, and extending the exhalation period to increase the retention of nicotine 12 . Compensatory behavior has the potential to render alternative tobacco products, such as low-yield cigarettes, ineffective at reducing risk of exposure or adverse health effects.
Respiration topography may also be used to study the effects of product characteristics on usage behavior due to variations in flow path restriction. Research has identified two main smoking patterns: mouth-to-lung (MTL) 13 and direct-to-lung (DTL) 14,15 . MTL is typically associated with cigarette use. One possible reason for this is due to the restrictive nature of the cigarette flow path which makes deep inhalations difficult. In contrast, we expect to see more DTL behavior in users of lower-resistance tobacco products, such as hookah and certain e-cigs like the JUUL. For any fixed inhaled volume, the concentration of nicotine is likely higher in DTL than in MTL since in DTL, the volume is comprised entirely of emissions from the tobacco product whereas in MTL, it is a smaller volume of puffed emissions that is then diluted in a larger volume of clean air during inhalation. Given this, it is conceivable that tobacco manufacturers are designing their products to have lower resistance, as is the trend from cig-a-likes, to sub-ohm, and finally to pod-style e-cigs, so as to facilitate DTL behavior and thereby increase nicotine consumption. This is, however, unverified but with a method to measure respirotry topography, we may in the future be able to link product characteristics (e.g. flow path resistance) to product use behaviors (e.g. DTL vs MTL) and their consequent health effects. By observing when inhalation occurs in relation to the puff, the full topographic smoking pattern can be obtained.
This paper aims to assess the feasibility of adapting the Hexoskin for ambulatory respiration topography observation and address some of the challenges, in particular 1) calibration of the Hexoskin Smart Garment, 2) calculation of respiration topography parameters observed from smokers in their natural environment, and 3) acceptability of the Hexoskin by the participants.

Human Subject Study Protocol
The study protocol consisted of an online participant recruitment and pre-screening survey (Table 1) prior to the intake meeting, a final screening and verification of eligibility according to the Inclusion/Exclusion Criteria during the intake meeting, the review and signing of the informed consent form, the deployment of the WRM for a 3-day observation period with an in-lab WRM calibration on each observation day, an exit questionnaire assessing the acceptability of the monitor and the study protocol, and finally the incentives. The study protocol was reviewed and approved by the Rochester Institute of Technology (RIT) Human Subjects Research Office Institutional Review Board (IRB). How long ago did you start using a hookah? Text entry * Question only presented if participant answered "Everyday" or "Some days" in Q1 ^ Question only presented if participant answered "Everyday" or "Some days" in Q2 Cigarette and hookah users were recruited from the RIT community via mass email and flyers between July 2019 to March 2020, when the study was halted due to the COVID-19 pandemic. To satisfy the Inclusion Criteria, the participant must: 1) Be of legal smoking age, 2) Answer "Some days" or "Everyday" to the question Q1 or Q2 in Table 1, and 3) Indicate that they had their own tobacco product since the study would not provide them with a product to use. We complied with the change in legal smoking age from 18 to 21 on 20 th of December 2019 due to the "Tobacco 21" legislation 16 . The Exclusion Criteria excluded participants with underlying cardiovascular and pulmonary diseases, pregnant women, or women intending to become pregnant from being eligible.
The observation period spanned three days. On the first day, prospective participants were invited to the laboratory for a final eligibility screening and were enrolled once they reviewed and signed the informed consent form. Enrolled participants were provided with the WRM along with instructions on how to use it. The participants were informed that they must wear the WRM whenever they are using the tobacco product type they were enrolled to use and were not required to wear the WRM when not smoking. The participants were also provided with a daily study log to self-report their use behaviour and compliance to the study protocol. At the end of the intake meeting, a WRM calibration was conducted on the participant and then they were allowed to leave and use their tobacco product freely. On the second day, the participants returned to the laboratory at a pre-scheduled time for another WRM calibration. The participants were free to use their desired product prior to and after the scheduled meeting. On the third day, the participants retuned to the laboratory and performed a final calibration. The participants were asked about the acceptability of the WRM and the study protocol in a structured interview. Finally, participants were provided a $25 incentive upon successful completion of the study protocol. Participants who dropped out of the study prematurely were provided with a $5 incentive.

Wearable Respiratory Monitor and Calibration
The WRM used in this study is the Hexoskin Smart Garment with the Hexoskin Smart Device (datalogger). The feasibility of using the Hexoskin for ambulatory measurement of respiration topography was previously assessed. The garment has two embedded respiratory inductance plethysmograph (RIP) sensors that measure the change in crosssectional area at the thorax and abdomen. The datalogger collects data from these sensors at a fixed rate of 128 Hz and stores the data internally. Data collection begins immediately once the datalogger is attached to the garment and continues until it is detached from the garment or runs out of battery. At the end of the study, the data is extracted from the datalogger via USB.
The two chest motion waveforms, measured at the thorax (TC) and the abdomen (AB), are related to the lung volume waveform as per the model (reproduced in Equation 1) introduced by Konno and Mead 17 . The lung volume waveform calculated using this model is an estimate of the actual lung volume and is denoted by ̂. The parameters in curly braces denote vectors with a length of N elements. The Volume-Motion (V-M) parameters, Ktc and Kab, convert the TC and AB waveforms from arbitrary units (i.e. counts) to volume units (i.e. mL). These V-M parameters are obtained via calibration of the WRM, which must be done on the person wearing the WRM to account for variations in physiology and breathing style across participants, and variations in respiratory sensor response across shirts.
During calibration, participants were instructed to breathe exclusively through a spirometer (Vernier, Beaverton, OR, USA) while wearing the WRM and a nose clip, and perform the following breathing exercise while seated: 5 normal breaths, followed by a deep inhale, a breath-hold, then an exhale, followed by 5 more normal breaths. This was repeated 3 times without pausing in between. Participants were instructed to abort the breathing exercise at any time if they experienced any discomfort.
The spirometer measures the actual lung volume waveform (SP) and is considered the true condition. By using Equation 1, the ̂ waveform can be calculated from the TC and AB waveforms and an arbitrary set of Ktc and Kab. The difference between ̂ and SP is defined as the residual ( , Equation 2).
The residual is calculated at each value of the SP, TC, and AB waveforms and is therefore itself a waveform with a length of N elements. For a given set of SP, TC, and AB waveforms, is a function of Ktc and Kab. The objective of the calibration is to find a pair of Ktc and Kab that minimizes . For this, the mean absolute residual (, Equation 3) is considered and is plotted as a function of Ktc and Kab to form a mean absolute residual surface. The ̃ is a scalar with units of [mL] and represents the mean of the absolute difference between the volume estimate and the true condition.
As per the study protocol each participant performed 3 calibrations, one on each day of observation. As a result, there were 3 mean absolute residual surfaces for each participant. To find the set of Ktc and Kab that yields comparable ̃ when applied to all 3 calibrations, we proposed finding the centroid to the polygon formed by connecting the minimum point of each surface. The values of Ktc and Kab at the centroid were then used to estimate the lung volume waveform (Equation 1) from TC and AB obtained while the participant used their tobacco product in the natural environment. This method of obtaining the values of Ktc and Kab is broadly applicable for N≥3 calibrations.

Respiration Topography Analysis
The data analysis was conducted using an added feature in RIT's previously described Topography Analysis Program (TAP) 18 . First, the TC and AB timeseries waveforms collected from the WRM were segmented into data for calibration, obtained during the calibration process, and data for calculating respiration topography, obtained during the natural environment observation. Both TC and AB waveforms were smoothed using a Savitzky-Golay 128 sample rolling average filter. TAP then identified the minima in each waveform and then performed baseline compensation by first generating the baseline trend waveform for both TC and AB by piecewise interpolation between the minima on each waveform. The baseline trend waveform of each chest motion waveform is then subtracted from the original waveforms. The resultant baseline compensated waveform would then nominally have minima at 0. The algorithm then identified the maxima locations in each of the baseline compensated waveforms. The Ktc and Kab parameters were then obtained from the calibration data segment using the method described above. These parameters were then applied to Equation 1 along with the TC and AB from the natural environment data segment to obtain the volume estimate waveform.
The volume waveform was then segmented into individual breathing cycles by using the previously found extrema locations. The minima at the start of a rising wave is the start of an inhale (and the start of the cycle) and the minima at the end of falling wave is the end of an exhale (and the end of the cycle). By considering the maxima, each breathing cycle can then be further segmented into an inhale, breath-hold, and exhale. Between any two minima there is one maximum, which corresponds to the end of the inhale. The segment of the waveform from the preceding minimum to this maximum is considered the inhale portion of the cycle. Subsequently, the period from the end of the inhale to the following minima is considered the exhale portion of the cycle. If a period of stable volume (neither rising nor falling significantly) is present following the end of an inhale before the start of the exhale, then this is considered the breath-hold period. During normal breathing, the breath-hold period is typically negligible whereas breath-holding is likely present following smoke/vapor inhalation during tobacco use.

Assessing Compliance and Acceptability
The daily paper study log was given to each participant that allowed them to selfreport their product use during the study. The participant was instructed to report the number of cigarettes or hookah sessions they had each day along with the number of times they used their assigned product without wearing the WRM. By using this information, the participant's compliance was assessed. The study log also allowed the participant to track whether they charged the WRM daily, which is necessary to prevent loss of data from the WRM discharging fully the following day.
During the exit appointment, each participant was given a questionnaire that assessed the acceptability of the study and the WRM. The questionnaire was administered as an interview, with the research administrator asking the questions and noting down the verbal responses from the participant.

Study Cohort
A total of 47 individuals started the pre-screening survey (Figure 1), of whom 19 did not complete the survey. Of the 28 individuals who completed the survey, 15 were found to be ineligible. Reasons for ineligibility included not being above the legal smoking age or not being a cigarette or hookah user. Of the 13 eligible individuals, 10 responded to the follow up email and scheduled an appointment. All 10 individuals were confirmed to be eligible during the intake appointment and were enrolled. All 10 enrolled participants were RIT students between the age of 21 and 29, of whom 5 were male and 5 were female. Of these, 3 did not complete the study protocol: 2 were voluntary dropouts and 1 was dismissed because they were non-compliant. Data from these 3 participants were not included in the analysis. Of the remaining 7 participants, 4 completed the study as hookah users, 2 as cigarette users, and 1 (PS6-02) completed the study twice, once as a hookah user and once as a cigarette user. As such, there are 8 sets of data (N=5 Hookah and N=3 Cigarette) presented for the waveform results and respiratory topography results. Enrolled participants' responses to the pre-screening survey are shown in Table 2.

Waveform Results
All 8 data sets from 7 participants were analyzed. Each data set included 3 calibration sessions and at least one session of tobacco use observation in natural environment. All hookah participants only had one tobacco use session, but all cigarette participants had more than one. A session is defined as the time from when the Hexoskin datalogger is plugged in to the shirt until it is detached. The Hexoskin datalogger only collects data while it is attached to the shirt and creates a new data file every time it is connected to the shirt. This section presents the results of the analysis from the calibration process through to obtaining the volume estimate waveform using one participant's data (PS6-11) as an exemplar case (Figure 2). Similar results were obtained across all data sets.   Figure 2a is the spirometer volume waveform [mL]. The bottom two panels in Figure 2a are the corresponding uncalibrated TC and AB waveforms [counts], respectively. The spirometer volume was calculated as the integral of the measured flow rate. Figure 2.b shows the mean absolute residual surface generated from the SP, TC, and AB waveforms from Figure 2.a, with Ktc and Kab each ranging from -50 to +50. The mean absolute residual surface shows that r ̃ is more sensitive to Ktc than Kab and that there is a region of minima represented by the valley in the 3D surface. The black line represents the line of fit to the minima region, effectively the 2D minima line to the 3D surface. The pair of Ktc and Kab that minimizes the surface is found by finding the location of the smallest value of r ̃ along this line. For the set of data shown in Figure 2, the values of Ktc and Kab corresponding to the minima of the surface were found to be 7.17 and 13.84, respectively. Figure 2.c shows the minima lines from the 3 calibrations conducted with participant PS6-11. The circular markers indicate the global minimum point on each line of local minima. The x marker indicates the location of the centroid of the polygon formed from the circular markers at the vertices. For the set of calibrations shown in Figure 2, the values of Ktc and Kab at the centroid were found to be 14.67 and 13.52, respectively. Exemplar TC and AB waveforms for participant PS6-11 are shown in Figure 3 for a single minute of one hookah session captured in the natural environment. As can be seen in the figure, the V ̂ waveform captured a mix of normal breaths (volumes of around 500 [mL]), deep breaths (volumes of around 1000 [mL] and greater), and breath-holds.

Respiration Topography
The computed respiration topography data of the 5 hookah data sets and 3 cigarette data sets are reported in Table 3. A total of 4,733 breathing cycles were observed across all participants, of which 232 were from cigarette smokers and 4,501 were from hookah smokers. The number of breathing cycles from cigarette smokers make up around 5% of the total data. The low number of breathing cycles observed from cigarette smokers compared to hookah smokers is a direct consequence of the study protocol that only required participants to wear the shirts during tobacco use. Since cigarette sessions are in the order of minutes and a hookah session is in an order of hours, the total breaths observed from cigarette users are much less than that from hookah smokers. There are also more hookah data sets than there are cigarette data sets.  Figure 4 shows exemplar descriptive statistics collected for participant PS6-04. The histograms in Figure 4show 3 characteristics: 1) The inhalation and exhalation volume distributions (Figure 4a) are overlapping, indicating that there is conservation of mass across the spectrum of breaths observed, which is in contrast to 2) The duration ( Figure  4b) and the flow rate distributions (Figure 4c) both show a distinct but consistent bias between inhalation and exhalation, with inhales being shorter and sharper and exhales being longer and more relaxed. 3) The volume distribution (Figure 4a) is right-skewed, with the mode (of around 200 [mL]) smaller than the mean (of around 300 [mL]), which is indicative of higher frequency of smaller volume breaths than deeper breaths. The duration and the flow rate distributions also appear to be right skewed but this behavior is not seen consistently across all participants.

Figure 4 -Exemplar respiration topography results from a participant (PS6-04, N=1229 breathing cycles observed). Top row -Distribution of inhalation and exhalation a) volumes, b) durations, and c) flow rates. Bottom row -Breath-by-breath comparison of inhalation to exhalation d) volumes, e) durations, and f) flow rates.
The grand mean of the ratio between the inhale volume to the corresponding exhale volume within a breath across all breaths captured across all participants is 1.04. A volume ratio close to 1 is indicative of the mass conservation on a breath-by-breath basis. This is visually illustrated by the scatter points clustering around the 1-to-1 line in the scatter plot of inhalation versus exhalation volume in Figure 4d. To satisfy the conservation of mass, the inhale volume should be the same as the exhale volume. However, there is some potential for variability in small time windows, e.g. on a breath-by-breath basis, due to the variability in breathing behavior and the ability for a person to tap into their lung's reserve capacity. This may explain some of the scatter points that lie farther away from the 1-to-1 line. It is expected that the conservation of mass should hold over time, which can be assessed by comparing the cumulative inhale volume to the cumulative exhale volume. The ratio of cumulative inhale volume to the cumulative exhale volume, summed across all participants, is found to be 0.996.
The grand mean of the ratio between the inhale duration to the corresponding exhale duration within a breath across all breaths captured across all participants is 0.654. This ratio being smaller than 1 is indicative of the bias towards the exhale being longer than the inhale. This is visually illustrated by the larger amount of scatter points above the 1to-1 line in Figure 4e.
The grand mean of the ratio between the inhale flow rate to the corresponding exhale flow rate within a breath across all breaths captured across all participants is 1.32. This ratio being larger than 1 is indicative of the bias towards sharper inhales than exhales. Since the flow rate is calculated from the ratio of the volume to the duration, this flow rate bias is inversely related to the duration bias. This is visually illustrated by the larger amount of scatter points below the 1-to-1 line in Figure 4f.
The consistency between inhale and exhale volumes and the bias between inhale and exhale durations and flow rates are likewise observed in all participants ( Figure 5). Figure 5a shows the mean and its 95% confidence interval ofinhale and exhale volumes across all participants. The confidence intervals are larger in the cigarette cases than the hookah cases due to the smaller number of breathing cycles observed in the cigarette cases. Blue, red, and black colors denote inhale, exhale, and cycle, respectively. Circular markers indicate hookah participants and square markers indicate cigarette participants. Figure 5b shows the mean and its 95% confidence interval of inhale, exhale, and cycle durations across all participants. In all participants, the mean exhale duration is longer than the mean inhale duration. In the absence of breath-holding, the cycle duration is the sum of the inhale and exhale durations. Figure 5c shows the mean and its 95% confidence interval of inhale and exhale flow rates for all participants. The mean inhale flow rate was higher than the mean exhale flow rate for every participant. This flow rate bias is consistent with the opposite bias in inhale and exhale durations.

Compliance and Acceptance of the Monitor
All enrolled participants (N=7) took the exit questionnaire and indicated that the Hexoskin shirt did not change their typical tobacco use behavior. However, one participant indicated that wearing the shirt made them feel like they were being monitored. Two participants indicated the shirt was 'tight', while 5 indicated it was either 'comfortable' or 'fine.' All participants reported wearing the shirt for at least an hour, with two reporting wearing it for 4-5 hours, and one reported wearing it for the whole day. All but one participant indicated that they wore the shirt only for the duration of smoking. All participants connected the datalogger only while they were smoking. All participants indicated that they felt comfortable wearing the shirt in a public setting. Four participants indicated the requirement to wear the shirt had no impact on their desire to smoke but the other three indicated it was either slightly discouraging or hard to remember to put on. All but one participant indicated that the shirt had no effect on their respiratory effort to puff their tobacco product. One participant indicated that the shirt had a minor impact on their ability to perform their regular non-smoking activities. When asked for the maximum amount of time they would be willing to wear the Hexoskin shirt, 2 participants answered that they would be willing to wear it for less than 3 hours, 2 participants answered up to 7 hours, and 3 said they would be willing to wear it for a day or more.

Discussion
Many calibration techniques have been introduced in the literature 17, [19][20][21] . We extended upon previous methods by analyzing the sensitivity of the residual to each V-M parameter. We saw that the mean absolute residual surface (Figure 2b) showed a significantly stronger sensitivity to Ktc than to Kab. This feature was present in all participants. A simple linear regression algorithm would have just returned a single pair of Ktc and Kab that was the "good fit" but we see that there are potentially an infinite number of combinations of Ktc and Kab with residuals within close proximity of each other (as illustrated by the valley in the mean absolute residual surface in Figure 2b). In the absence of more information, it is not possible to determine the pair of Ktc and Kab that would qualify as a "good fit" beyond just a value judgement on the part of the analyst. One potential way to facilitate this is by obtaining more information about the physiology and the breathing style of the participant. Such as by calibrating under different body postures or breathing maneuvers and patterns. We believe that the method presented here is a reasonable first step towards a more robust way to derive Ktc and Kab than the linear regression method presented in the past.
Additionally, we prototyped the logic to derive a single pair of Ktc and Kab from multiple calibrations (N=3 in this paper) using the centroid method. Other approaches considered included: i) Taking the mean Ktc and the mean Kab of the N calibrations, ii) Only use a pair of Ktc and Kab from one of the N calibrations, iii) Apply the pair of Ktc and Kab from the most recent calibration for the data collected following the calibration leading up to the next calibration. iv) Combine all N calibrations into a single data set to determine a single pair of Ktc and Kab. It is presently unclear what the best method is from these, but we plan on investigating this further.
The calibrations were conducted while the participants were seated only, with their torso perpendicular to the ground, which we deemed to be the posture the participants would most likely use during their tobacco use sessions. Some prior research suggests that a participant's V-M parameters may vary based of their body position 22 . Hence, it is possible that the inhalation and exhalation volumes measured by the WRM calibrated while seated may not be accurate if the participants were not seated while smoking. In a future study, we plan to investigate the impact of calibration posture on the accuracy of the volume measurements. The duration based respiration topography parameters, such as inhale and exhale durations, are not impacted by the magnitude of the volume waveform and therefore not impacted by the calibration since they are calculated from the temporal locations of the extrema on the lung volume waveform.
We were able to deploy the monitor for a 3-day observation study, which covered at least one full 24-hour period. We were also able to analyze all the data collected from this period of observation and obtain both lung volume waveform and respiration topography for all 7 participants. For this, we developed a computer algorithm to: i) Pre-process the data from the Hexoskin, ii) Perform the calibration and apply it to the measured TC and AB waveforms to obtain the lung volume waveform, and iii) Derive the respiration topography. This was necessary because the Hexoskin did not come with the necessary software to obtain either lung volume or respiration topography 1 . We did not modify the Hexoskin hardware in any way for this study.
The data presented here are from participant's smoking session and are comprised of two predominant underlying behaviors: natural tidal breathing and smoking related ventilation, i.e. inhalation of tobacco emissions post puffing with potential breath-holding prior to exhalation. Although we have not attempted to discriminate between the two behaviors in this paper, the distribution (Figure 4) of larger amounts of sub 1000 [mL] volumes and smaller amounts of volumes greater than 1000 [mL] is indicative of natural tidal breaths and smoking related ventilation, respectively. The ability to discriminate between the two behaviors is necessary for identifying smoking specific respiration topography. This is currently a focus of our future development. For this, we will leverage our wPUM topography monitors 23 to identify the start and end of a puff relative to the lung volume waveform measured by the WRM. This will help in discriminating the breaths that are associated with smoking from the natural tidal breaths. Additionally, this will potentially allow us to discriminate between MTL and DTL. The characterization of these smoking patterns are also topics for future research.
It was observed that in all participants, the mean exhale duration was longer than the mean inhale duration. This is consistent with the accepted pattern of tidal breathing in healthy adults. We predict that this phenomenon is also present in smoking related ventilation cycles, as a mechanism for increasing the nicotine uptake by extending the smoke retention time. Additionally, changes in the duration of either the inhale or the exhale, or the ratio between the two may, also be an effect of compensatory behavior. As expected from the relationship between the exhale duration and inhale duration and the consistent volume inhaled and exhaled, the data presented here indicate that the mean inhale flow rate is higher than the mean exhale flow rate.
The extrema locations in the volume waveform has a strong impact on the respiration topography since every respiration topography parameter is calculated based on the location of the extrema that prescribes each breathing cycle. Therefore, the algorithm used to find these extrema points have an impact, since any misplaced extrema will affect the respiration topography of that breathing cycle and possibly those of the adjacent breathing cycles. The quality of the data and the nature of the behavior being observed also has an impact on the algorithm's ability to obtain respiration topography, including irregularities in the volume waveform that may be a result of behaviours such as walking, talking, and yawning, or sensor slippage on the body. More work still needs to be done to assess the sensitivity of respiration topography to these irregularities in the volume waveform. The current algorithm operates on the volume waveform directly and uses the prominence values of the peaks and troughs of the breathing cycles to find the extrema, but alternative approaches exists that make use of frequency domain analysis or machine learning. More work must be done and presented to make a meaningful comparison between methods of obtaining respiration topography from lung waveform.
The participants' acceptability of the Hexoskin is key to assessing its viability as a tool for ambulatory measurement of respiration topography. Prior to the start of the study, we were concerned that the participants would not be willing to wear the Hexoskin shirt for extended periods of time, i.e. to cover the whole day of observation. Therefore, we only required the participant to wear the shirt while they were using their tobacco product and we expected that the participants would take the shirt off in between sessions. From the exit interview results we can see that the majority did remove the shirt in between sessions but most indicated that they would be willing to wear it for extended periods of time. In a future study we intend to recommend the participant wear the shirt for a full day of observation. Data collected outside of their smoking session will help establish the participant's baseline tidal breathing behavior and help with identifying smoking related ventilation. The downside to the requirement to wearing the shirt of an extended amount of time is that it we would lose the natural indication of the start and end of a smoking session that came with the participant attaching and detaching the datalogger from the shirt. Another option is for the participant to wear the shirt throughout the day but detach the datalogger between sessions. The disadvantage here is that there is a risk that the participant would forget to reattach the datalogger prior to starting a session, causing the loss of valuable data. Alternatively, we can leverage our wPUM topography monitor to provide the start and end time of each puff, and the start and end time of each session.
Another concern that we had was in the amount of data that would be collected if the shirt was worn for an extended time. The Hexoskin samples each chest motion sensor at 128 [Hz], which after 12 hours would result in over 5 million samples. In addition to the storage concerns, we anticipated computational difficulty with processing large vectors and finding extrema locations in large sets of data. In practice thus far, we have not encountered any issues with our algorithm in terms of the data size. We expect that the algorithm would be scalable to larger datasets obtained from longer observation periods.

Conclusions
A method for calibrating a commercially WRM across multiple observation days was presented. The thoracic and the abdominal signals were measured using the Hexoskin Smart Garment on 7 participants while they were using their tobacco product. Lung volume waveforms and respiration topography parameters were obtained using an extension of RIT's Topography Analysis Program. The acceptability of the Hexoskin was assessed, with most participants indicating willingness to wear the shirt for an extended amount of time, allowing for longer periods of observation. This is the first study to demonstrate ambulatory measurement of lung volume in the natural environment. The ability to quantify respiration topography, combined with existing techniques of capturing natural environment puffing topography, will lead to a comprehensive objective assessment of tobacco use behavior.

Institutional Review Board Statement:
The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board of the Rochester Institute of Technology (Project Title: Pilot Multiproduct Topography Data, approved June 26, 2018).

Informed Consent Statement: consent was obtained from all subjects involved in the study
Data Availability Statement: The data presented are contained within the article