Ultra fit mask fabrication

Materials used in the UFM are readily available and inexpensive. For the base mask, a 3PM with 3 pleats was selected (SupplyAID, SnowJoe LLC, Hoboken, NJ) due to its availability. This 3PM is compliant with GB/T32610-2016. However, any rectangularly-shaped pleated face mask and medical-use surgical mask can be used as a base mask. As the supply of face mask improved and the ASTM F3502 standard emerged, a base mask compliant to NIOSH filtration test was used. Also, the dimensions of base mask were carefully determined by analyzing non-medical 3PM. A base mask with 175 x 95mm (pleated, as is), 175 x 160mm (unpleated), ear loop string lengths of 180-190mm was used for a broader population. In addition, based on surveying spring constants of elastic ear loop strings from 9 mask manufacturers, we used strings with less than 12 N/m at 50 grams. Strings with >12 N/m cause discomfort and pain due to more pressure exerted on the back of the ears. Four segments of Zinc-galvanized steel wires with a diameter of 0.023” (8872K16, McMaster-Carr, Elmhurst, IL) were incorporated on the four edges on the rear side of the base mask, which were then taped over with medical adhesive tape for soft skin contact (MedFix^™, Medline, Northfield, IL). Empiric experimentation across a range of different wire diameters demonstrated the best performance with 0.50mm to 0.58mm bare wire, where experimentation was to deform a Ultra fit mask by placing it on the experimenter’s face and skin. The material of the inner flap layer was spunbond polypropylene nonwoven fabric with a weight of 30–40 grams per sq. meter (GSM), the identical material as the inner mask layer, which is a low-cost, non-hazardous material for skin contact. A 30–40 GSM fabric was chosen because of its reliable ultrasonic weldability. We observed sonic welding with a lower GSM fabric resulted in weak joining. Custom-designed ear loop adjusting components were cut from a stock polypropylene sheet (Formex-GK17, ITW, Carol Stream, IL) and bonded with hot-melt glue on the upper corners on the front side of the base mask. The bulk cost of materials is estimated to be less than $0.05 per mask. At the time of writing, we demonstrated that these modifications can also be performed with a single ultrasonic welding operation, eliminating the need for adhesives or tapes.

Visualization of face seal leakage

For visualization of exhaled vapor, a custom laser particle scattering setup was devised. An e-cigarette was used to produce vapor particles with a mean particle size ranging from 174 to 236 nm [16]. Before donning a mask, the user inhaled vapor, donned the mask, and then exhaled. Particles were assessed using light scattered from a green laser pointer (Laser 303, λ_c = 532nm, P_max <5mW) equipped with a cylindrical lens (LK1395L1, Thorlabs, Inc., Newton, NJ) to create a linear beam. The beam illuminated sagittally toward the subject and coronally toward the chin. The exhalation event was recorded on an Apple iPhone X mounted on a tripod. For relative scattering analysis, the green channel of video files was analyzed over time using MATLAB. Relative scattering was calculated as a sum of pixel values divided by the segmented area. Plots were then normalized to the theoretical maximum. Also, the participant in Fig 2 has given written informed consent (as outlined in PLOS consent form) to publish the case details.

For infrared thermal imaging, two cameras, T440 (Teledyne FLIR, Wilsonville, OR) and CompactPro (Seek Thermal, Santa Barbara, CA) were initially used. For convenience of image transfer, CompactPro was used for image acquisition. The room temperature was maintained around 27 °C. The thermal camera was positioned about 1m away from the subject such that the subject’s head fills >50% of the image. The subject, sitting on a chair, breathes through the nose at rest for at least 2 minutes. For presentation purpose, images were cropped around the subject’s head. For plotting thermal variation over time, 5-by-5 pixel regions were segmented for the left and the right infraorbital regions and a 20-by-20-pixel region was segmented for the center of the mask (Fig 3A). Mean values of the segmented regions were plotted over time. Thermal video data was acquired at 8.6 frames/second.

An aerosol dye challenge was used to evaluate the protection provided by the mask to external aerosol exposure. The aerosol dye was generated 20 cm away from a head model, which was 3D printed using data obtained from the NIOSH Anthropometric Data and ISO Digital Headform [17], and was modified to outfit with a 2 cm tube connected to the mouth. To simulate inhalation and exhalation, a 2L resuscitator bag (Vodeson) was modified by closing the one-way valves on the back of the bag and removing the valve at the interface with the mouthpiece. The resuscitator bag was physically compressed and then allowed to expand in 3 second cycles over a 30 second test period to simulate breathing. Aerosol particles (27.94 ± 18.96 μm in diameter, 483,000 ± 102,000 particles per second, mean ±SD) of fluorescent dye (EC6 RECOLOUR Dye UV Green, XSPC) were generated from a home-use ultrasonic mesh nebulizer (Model MN-01A). Dye on the surface of the mannequin was imaged using a 365 nm ultraviolet light source (Model UVGL-25, UVP, LLC, Upland, CA) and a digital camera. The percent of the mannequin head’s surface coated in dye underneath the face mask was determined using image analysis in ImageJ 1.53E (National Institute of Health).

Quantitative fit test

For a quantitative fit test, a commercial PortaCount^® Respirator Fit Tester 8038 was rented (Raeco Rents, LLC, Gurnee, IL) for the duration of experimentation. Automatic daily checks were performed before use. A standard protocol per 29 CFR 1910.134, which consisted of 8 exercises, was employed. After five exercises comprising different breathing styles (normal and heavy), head motions (side-to-side and up-and-down), and reading a provided passage, a wearer was asked to grimace to intentionally break a face seal, followed by bending over towards the toes before the final normal breathing was attempted. A fit factor of each exercise was automatically computed and displayed at the end of each exercise except for non-breathing ‘grimace’ exercise. An overall fit factor was automatically computed at the end of the whole exercise cycle. With N95 Companion mode enabled, PortaCount^® 8038 discriminates the contribution of leakage through the filter by use of electrostatic classifier and counts only particle sizes in the range of 25 nm to 60 nm. Therefore, the fit factors, as a measure of face seal leakage, are the ratio of ambient particle concentration outside the mask to that inside the mask for particle sizes between 25 nm and 60 nm. A conventional 3PM disposable mask without modifications and KF94 mask (AirQueen Nano, Soomlabs, Korea), a Korean standard rated to filter 94% of particles larger than 0.3 microns in size [18], were used as reference controls (S2 Fig). The 3PM used in our experiment was sold as SupplyAID by SnowJoe LLC, which also served as the base mask for UFM. A total of thirteen volunteers participated in the quantitative fit test.

User validation

An online survey form was created and circulated for collecting user feedback. Questions asked in the survey were multiple-choice questions and short questions about what type of face mask a user wears, and the user’s assessment of the UFM. Results are available in the supplementary information.

Modeling airborne infection risk mitigation

To evaluate how the design modification to the UFM would influence the risk of airborne pathogen spread we used the computational model developed by Peng et al. [19, 20]. Briefly, the model uses the concept of an enclosed box to estimate the amount of aerosol contamination taking into consideration: ventilation, volume of the space, and the amount of virus put into the air by infected individuals (factoring in mask leakage). The model then calculates an airborne risk parameter that considers the amount of virus in the air over the exposure period, the volume of air uninfected individuals takes in during breathing, and mask efficiency. The Wells-Riley model of infection is then used to calculate the probability of an individual becoming infected after exposure [21]. We derived the mask efficiency for the UFM based upon the results of the quantitative fit testing, filter testing, and the aerosol exposure test. For the cloth mask (mask efficiency 20%), excellent cloth (mask efficiency 50%), and N95 (mask efficiency 95%), reported mask efficiency values were used [22]. The mask efficiency values entered into the model generated in Peng et al. and the fraction of people with masks was varied systematically. The baseline settings for ventilation, volume, duration of stay, activity adjusted breathing rates, and infectious quanta for COVID-19 within the classroom and subway environments, were used. The number of individuals in the classroom and subway car were set to 20 and 35, respectively, and there was assumed to be one infective individual in each case. Data in the classroom was normalized against the risk of a secondary infection occurring in the absence of any masks over a 50-minute class period. As the duration of ride on a subway is highly variable, we calculated the absolute infection risk parameter, which is independent of the duration of the ride, at 0, 25, 50, 75, and 100% mask utilization.

Statistical analysis

Data processing and statistical analysis were performed using a combination of MATLAB, ImageJ, and GraphPad Prism 9.0. All graphs and reported values represent mean ± standard deviation unless otherwise stated. A Shapiro-Wilks test was used to test if data were normally distributed before selecting parametric or nonparametric tests. Comparisons across multiple groups were performed on nonparametric data using Kruskal-Wallis for multiple groups and Mann-Whitney for paired tests. For parametric data, a one-way ANOVA was used. A Tukey post-hoc test was applied to correct for multiple comparisons and an adjusted p-value of < 0.05 was used as the threshold for significance. The specific test used is described in the caption of each figure.

Ethics statement

Our quantitative fit test protocol was exempted by the Mass General Brigham Institutional Review Board (#2021P001871) as the protocol meets the criteria for exemption 45 CFR 46.101(b)(#). Also, the participant in Fig 2 has given written informed consent (as outlined in PLOS consent form) to publish the case details.

Ultra fit mask—Improved mask design

To facilitate improved fit and seal gaps, we added a moldable wire element along the entire periphery of a conventional 3-ply disposable mask (3PM) (Fig 1A and 1B). With the moldable element, the mask gains a pliable structure that can conform to any face, with minimal gaps (S1 Movie). Compared to the 100 mm-long nose bridge wire found in non-medical 3PM, a longer and thicker nose wire (length > 120mm, diameter > 0.58mm) provides a better seal across the entire nasal-infraorbital region (bizygomatic breadth). The wires on the cheek and the chin edges of the mask provide a fitting experience similar to cup-shaped respirators and enable customized shaping of the mask edge. Pinching the side or chin wires allows to tighten the fit and further reduces gaps (S1 Fig). As such, these face-conformal adjustments improve fit, particularly for users with smaller faces.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Ultra fit mask for reducing face seal leakage on the periphery of the mask body.

(A) Ultra Fit Mask (UFM) fitted on a head form compliant to ISO 16976–2 Respiratory Protective Devices (medium size). (B) Exterior and interior schematics of prototype UFM. (C) Additional cough-trapping inner flap layer attached on the interior side is highlighted in blue.

https://doi.org/10.1371/journal.pone.0281050.g001

To further promote a tight face seal, we added adjustable tension elements to the mask body. In conventional 3PM, elastic ear loop strings compress the mask onto the wearer’s face. The compressive force on a wearer’s face and the back of the ears is broadly a function of wearer head size or, precisely, the bitragion-subnasal arc (BSA) length, and the length and elasticity of the ear loop strings. Therefore, users with a longer BSA length suffer from irritation due to high pressure on the back of the ears while having a tight face seal, whereas other users with a shorter BSA length have a loose face seal without proper ear loop adjustment. To accommodate a broader population, in UFM, a longer ear loop strings were used, 180-190mm, which is 10-30mm longer than the ones found in 3PM. Also, elastic ear loops strings with a spring constant less than 12 N/m at a mass of 50g were used. Furthermore, a pair of tension adjustment components was integrated onto the base mask that cinches the elastic ear loop strings and allows length and tension adjustment by the users. Appropriate tensioning in combination with the face-conforming mask shape maximizes both comfort and face seal (Fig 1A).

To prevent the escape of droplets via the sides of the mask, we added an additional layer as a “cough trap” flap (Fig 1C). Coughing generates pressure that displaces the mask away from the face, reducing face seal exactly when it is needed most to capture rapidly ejected respiratory droplets. The incorporated inner flap serves to capture these droplets.

Qualitative experiments of mask fit

Laser particle scattering of exhaled vapor.

To evaluate the ability of the masks to reduce aerosol spread, we analyzed video recordings of vapor exhalation via laser light scattering as a qualitative assessment of mask fit. To image escaping vapor, video recordings were taken from the side and the front of a wearer (Fig 2A and 2C). Vapor escaping through open gaps for the 3PM was visible near the forehead, while vapor release through the side of the mask was clearly visible as a gray plume near the ears (Fig 2A, red arrow). Qualitative assessment of UFM vapor unequivocally demonstrated less vapor escape, with more vapor instead being forced through the mask filter material (Fig 2A, yellow arrow). These observations were confirmed by plotting relative laser scattering (RLS) of every frame of the acquired videos over time (Fig 2B and 2D). Leaked vapor from the UFM at the nasal-infraorbital side was nearly an order of magnitude less for the first 1.5 seconds than that with the 3PM (Fig 2B). After 1.5 seconds, a higher RLS with the UFM was observed, corresponding to scattering of floating vapor that penetrated the mask filter material earlier (S2 Movie). The UFM demonstrated negligible cheek leakage, with only a small leak plume arising from the chin region (Fig 2C, yellow arrow). With the UFM, the amount of laser scattering from around the cheeks due to leakage was reduced by ca. two orders of magnitude.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Evaluation of exhaled vapor using laser scattering.

(A, C) Side-view and front-view exhaled vapor imaging and histograms of the segmented area (white dotted box). Red arrows indicate particles that escaped the mask, while yellow arrows indicate particles that passed through the mask filter material. (B, D) Normalized laser scattering of the segmented forehead and lateral regions over an exhalation period.

https://doi.org/10.1371/journal.pone.0281050.g002

Infrared thermal imaging.

To further assess leakage sites for both inhalation and exhalation, we used infrared thermal imaging during normal breathing in an air-conditioned room (S3 Movie). Synchronous thermal changes at the mask surface with breathing cycles were visible as relatively cool inhaled air and relatively warm exhaled breath pass through the mask filter material. Inward and outward leakage along the nose-infraorbital edge of the mask were visualized as zones of high thermal variation over time (Fig 3A). The ingress of cool air through the nasal-infraorbital edge of the mask at the end of inhalation and egress of warm breath at the end of exhalation indicated leakage for the 3PM. For the UFM, the nasal-infraorbital edge remained relatively warm during breathing cycles except for a small ingress on the right side of the wearer (Fig 3A). Over a 2-minute breathing cycle, the UFM showed larger temperature variation in the center of the mask consistent with cycling of cool inhaled air and warm exhaled air traversing the mask filter material (Fig 3B) Thermal variation analysis at the right and left sides of the nasal-infraorbital edge of both masks showed that a sinusoidal response in the UFM was less pronounced compared to 3PM. A higher offset temperature around the edge of the UFM was due to increased skin contact owing to the reduction in mask edge gaps (Fig 3C). Overall, thermal imaging demonstrated that the UFM directed air through the filter material and reduced ingress and egress air leaks around the mask edge.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Infrared thermal imaging of mask surface over breathing cycles.

(A) Infrared thermal snapshots at the end of inhalation and exhalation with 3-ply mask (3PM) and UFM were obtained from video recordings of breathing cycles. Three subregions at the left and right side of the nose (red dotted boxes, 5 x 5 pixels), and the center of the mask (red dotted box, 20 x 20 pixels) were selected for thermal variation analysis over multiple respiration cycles. Thermal variation of the (B) center subregion and the (C) left and right side of the nose over a 120-second breathing cycle.

https://doi.org/10.1371/journal.pone.0281050.g003

Fluorescent aerosol dye challenge.

To assess the impact of improved mask fit on protection from external aerosols, the 3PM and UFM were challenged with fluorescent droplets (Fig 4). Exposure to aerosol droplets without any protection resulted in coverage of 77 ± 7% of the mannequin surface and 3PM masks reduced the dye-covered area by 22% (Fig 4B) (P < 0.05). Dye accumulation was observed in the same zones of poor fit identified during thermal imaging. This observation demonstrates that these areas of poor fit provide an opportunity for dangerous aerosols to bypass the mask filter material. The UFM was three times more effective at reducing aerosol ingress compared to the 3PM (Fig 4C) (P < 0.0001), even though the hard plastic of the mannequin reduced the ability of the UFM to form the tight seal normally generated as the mask is pulled against compliant skin by the adjustable tethering device.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Aerosol dye examination of 3PM and UFM.

(A) Schematic of the experimental setup where an aerosolized fluorescent dye was nebulized in the vicinity of the headform, with exhalation and inhalation breathing simulated by a modified resuscitator. (B) Fluorescent droplets deposited on a NIOSH medium headform with no mask, 3PM, and UFM. For scale, each image measures ca. 18 cm wide. A lower percentage of dye coverage indicates improved protection from external aerosol droplets. (C) Relative exposure of the mannequin to aerosols assessed by the relative area covered by aerosolized fluorescent dye. Results represent the mean ± SD of 6 repeated trials. * and **** indicate adjusted P < 0.05 and P < 0.0001 compared via a one-way ANOVA with Tukey post-hoc test.

https://doi.org/10.1371/journal.pone.0281050.g004

Quantitative fit test

We performed quantitative fit testing using a PortaCount^® fit tester. 13 volunteers underwent standard fit testing of 3PM, KF94 and UFM samples using the methods outlined in 29 CFR 1910.134, which is the standard minimal risk fit testing method used in non-research employment contexts. We considered this and the subsequent user surveys to be IRB exempt under categories 2 and 3 as defined in 45 CFR 46.101(b)(#). Subsequent IRB review of the testing protocol confirmed IRB exempt status. The average UFM fit factor (FF) was 12.9 ± 7.2 (Fig 5A), which was 320% of the 3PM FF (FF = 4.0 ± 1.8, P < 0.001) and 240% of the KF94 mask FF (FF = 5.5 ± 5.9, P < 0.001) respectively, while the observed difference between the 3PM and KF94 mask FFs was not statistically significant (P > 0.9999) (Fig 5A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Quantitative fit test results of 3PM, KF94 and UFM.

(A) Overall fit factors of quantitative fit test by PortaCount^® 8038 for n = 13 volunteers with 3PM, KF94 and UFM. (B) Fit factors between novice (n = 10) and experienced users (n = 3). (C) Effects of facial hair with 3PM and UFM groups. UFM with no hair outperformed both 3PM with and without hair and as well as UFM with hair. was used. (D) Exercise-specific fit factors measured with PortaCount^® 8038 of 3PM, KF94 and UFM. After reading a provided rainbow passage (‘Talking’), a non-measuring grimace exercise was performed per the OSHA fit test protocol. The post-grimace normal breathing exercise was labelled as ‘Normal Breathing 2’ to differentiate from the earlier normal breathing (‘Normal Breathing 1’). Bars are of the mean with error bars indicating the standard deviation. A Friedman test was performed for statistical analysis. *, **, ***, and **** indicate adjusted p-values of less than 0.05, 0.01, 0.001, and 0.0001, respectively.

https://doi.org/10.1371/journal.pone.0281050.g005

During the ‘Bending Over’ exercise, all three masks had a lower FF, and the UFM FF did not show a significant improvement compared to the other two masks (Fig 5D) (P > 0.05). Bending-over involved moving the upper body while a hose was connected to the mask during the test. Facial leakage was introduced due to the large body motion and strain created by the experimental setup in this position. Three experienced UFM users, whom as test users didn’t require donning instructions, scored an average overall FF of 23.7 ± 5.5 (P < 0.001) whereas novice UFM users scored an average overall FF of 4.7 ± 2.5, suggesting further face fit improvement with experience (Fig 5C). Interestingly, in novice users the UFM FF was 2.5 times higher than the 3PM FF (P < 0.005), consistent with a marked improvement in performance, even in novice users. While, in our user group, the presence of facial hair reduced fit factors, which was consistent with previous finding [23] (Fig 5D), the UFM FFs were still significantly higher compared to 3PM FFs in both with and without facial hair groups (P < 0.001 and P < 0.0001, respectively). Overall, the UFM group significantly outperformed the 3PM and the KF94 groups, regardless of familiarity, facial hair, or body movement.

User evaluation

To assess ease of donning and perceived comfort, we conducted a survey of individuals that had previously used the UFM. Each user had thus viewed the donning instructions and had experience using the UFM (S3A Fig). Most responded that the UFM was easier or similar in difficulty to don compared to conventional 3PM, with only 13% reporting more difficulty donning the UFM (S3B Fig). More than 90% of the respondents found the UFM similar or superior in comfortability compared to the 3PM disposable mask (S3C Fig). With appropriate instruction provided, ease-of-use for UFM was comparable to the 3PM.

Modeling UFM impact on community spread of an airborne pathogen

To evaluate the potential impact of the UFM on the risk of infection in the context of a classroom and mass transit, we used a model of aerosol disease transmission using parameters obtained from the current COVID-19 pandemic [19]. Masking was beneficial in the model no matter the quality of the mask, but as fit and filtration efficiency increased the amount of protection increased (Fig 6A). At with 50% of the class using masks the cloth mask, 3PM, excellent cloth mask, UFM, and N95 respirator reduced the relative risk of airborne infection by 18%, 24%, 44%, 62%, and 73% compared to no masks being used, respectively (Fig 6A). We calculated the relative impact of each mask type on the airborne infection risk parameter, a predictive measure of the rate an individual is exposed to an infectious agent, for a subway car. We found similarly that the UFM outperformed both model cloth masks and the 3PM (Fig 6B). At 50% mask prevalence the UFM use dropped the airborne infection risk by 50.4% compared to the 3PM. In both model situations, the UFM approached the efficacy of N95 respirators to prevent the community spread of COVID-19.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Model of airborne infection risk by mask type and usage.

(A) The relative risk of a new airborne infection occurring in a classroom type environment with varying rates of masking and mask types. (B) The calculated individual infection risk parameter for individuals riding a subway train, assuming the presence of an infective individual. The dashed purple line indicates the risk parameter for no masking.

https://doi.org/10.1371/journal.pone.0281050.g006