During the global pandemic of 2020, the diffusion of COVID-19 disease has caused more than 700 million cases and 7 million deaths worldwide [1]. The transmission of respiratory disease is based on three kinds of mechanisms [2]: (i) direct or indirect contact with mucous, skin, or contaminated surfaces; (ii) large droplet emission transmitted to the nose or mouth of the receiver; (iii) aerosol transmission of smaller particles and droplet nuclei, which can remain suspended for a long time. The infected droplet can be expelled by breathing, speaking, coughing, and sneezing. Among these respiratory activities, coughing was the most studied since it is more intense than breathing and speaking and could occur more frequently than sneezing [3]. In order to limit the spread of the SARS-CoV-2 virus, the World Health Organization (WHO) recommended wearing face masks. This kind of protective strategy has played a fundamental role in the reduction of infection rate [4,5].

In recent years, researchers have intensively studied the fluid dynamics of droplet emission during human respiratory activities, with particular attention to the cough. The flow field of saliva droplets and the effectiveness of protective face masks were evaluated both numerically [6–11] and experimentally [12–19]. These works highlight that the use of protective face masks significantly reduces the saliva droplet emission in front of the person. However, the flow can be partially leaked through gaps at the edge of the masks if they do not fit correctly on the face of the individual. The leakages are concentrated upwards, downwards, and on the side of the mask and could be very dangerous since they are not filtered by the face mask [3]. Furthermore, it was seen that the larger saliva droplet leakage is on the side of the mask, and it is directed towards the back of the person. Rearward leakage can be dangerous as it increases the risk of infection in people positioned behind the subject. This condition can be particularly critical in theaters, cinemas, schools, and public transportation.

Most of the experimental works applied PIV (Particle Image Velocimetry) [12–14] or the Schlieren [15–17] techniques for the quantitative evaluation of saliva droplet distribution. Qualitative flow visualizations were also performed by means of a smoke tracer and a laser source [18,19]. However, PIV and Schlieren techniques require a quite complex experimental setup. Furthermore, the use of a pulsating laser source and a tracer (PIV) could be dangerous for the living subject. Infrared thermography could be a solution to overcome these issues. In literature, only a few studies applied this technique to address the effectiveness of protective face masks [20–24]. Some studies [20,21] investigate the thermal comfort induced by wearing a protective face mask. They used the face skin temperature, measured by infrared thermography, as a key parameter to evaluate the subject’s thermal comfort. Du et al. [22] evaluate the air permeation capability of three types of face masks by coupling infrared thermography and X-ray tomographic imaging techniques. A warmer region in the infrared image was associated with high air filtration, while a colder zone was associated with low air filtration. The filtration performance of different types of face masks was also studied by Fragoso-Mandujano et al. [23]. In this study, infrared thermography was used to detect the temperature distribution induced on a screen placed in front of the test subject while breathing. Koroteeva and Shagiyanova [24] study the effect of wearing surgical, cloth, and N95 face masks on the exhaled airflow near a human face while coughing, talking, or breathing. A mid-wave infrared camera was adopted to measure the thermal distribution induced by the carbon dioxide concentration increase during exhalation.

This paper presents a novel non-invasive method to evaluate the saliva droplet leakage towards the back of a subject wearing a face mask while breathing and coughing. The proposed method adopts infrared thermography to measure the thermal footprint released by the saliva droplet on a screen placed behind the subject. Indeed, during exhalation, the particle cloud exits at a higher temperature than the screen, which was kept at 20∘C. When the cloud reaches the screen, a temperature increase could be detected by an infrared camera. Overall, the proposed approach provides a safe and non-invasive means to detect saliva droplet leakage behind the subject while offering a simpler and more accessible experimental setup compared to conventional techniques such as PIV or Schlieren imaging. The measurements were performed on two types of face masks (surgical and FFP2 masks), involving two subjects, one male and one female. In particular, the maximum temperature, the spread area, and the detection time of the thermal footprint were considered to quantitatively analyze the effects of the face masks. The maximum temperature is related to the intensity of the thermal footprint, which corresponds to the amount of saliva droplet emission behind the subject. The spread area deals with the diffusion of the saliva droplet, and it was evaluated by the mean temperature in a reference area. The detection time is the time interval between the cough event and the first peak in the time history of the temperature and provides an estimate of the droplet arrival time.

The paper is organized as follows. Section 2 describes the experimental apparatus, the image post-processing, and the test procedure. Section 3 presents the results of the experimental measurements. Finally, the conclusions are reported in Section 4.

Two types of protective face masks were tested, namely surgical and FFP2 masks. The face masks were worn by two subjects, one male (subject 1) and one female (subject 2), in order to verify the influence of face conformation on the saliva droplet thermal footprint. The two subjects have an age ranging between 25 and 35 years. The male subject has a beard, which would influence the saliva droplet leakage. The male subject also presents a better fit of the masks on the face. Informed consent was obtained prior to performing the measurement, including permission for publication of the images. The study was conducted in accordance with the Declaration of Helsinki. The subjects were seated on a chair placed at a distance of 10 cm from a black fabric screen located behind them. Each subject was required to perform two deep breaths followed by two coughs as a standard sequence for all the tests. During this sequence, the thermal footprint behind the subject was measured by infrared thermography. The quantitative analysis of the thermal footprint was performed excluding the emission zone near the subject’s neck. This decision reflects the consideration that emissions at head height are more relevant for infection risk than those near the neck. The tests were performed in an air-conditioned environment with the ambient temperature set to 20∘C. The temperature difference between the ambient and the exhaled airflow must be sufficiently large to ensure the detection of the thermal footprint on the screen. If the ambient temperature approaches the temperature of the exhaled flow, the resulting thermal contrast will be reduced, thereby hindering accurate detection of the thermal footprint. Therefore, the method presented in the paper is applicable if this condition is satisfied. The air-conditioning was stopped 10 minutes before the test started in order to avoid the influence of air circulation. Then, the thermal camera’s internal image correction was performed, and image acquisition began, and the subject performed the standard sequence. After the end of the recording, enough time was waited to restore the screen and environment temperature to 20°C. For each test, a new protective face mask was used.

The thermal camera adopted is the Flir A655sc. It is a long-wave bolometric thermal camera (wavelength ranging between 7.5 and 14 μm), with a NEDT (Noise Equivalent Differential Temperature) equal to 30 mK and a resolution of 640 × 480 pixels. The lens adopted has a field of view of 25∘ × 19∘ and a focal length equal to 24.6 mm. The thermal camera was placed at a distance of 1.5 m from the screen, resulting in a field of view of 66 × 50 cm and a single pixel dimension of 1 mm2. Fig. 1 shows the position of the subject and thermal camera with respect to the screen. Ambient temperature and relative humidity were measured before each test with an uncertainty of ±0.6∘C and ±2.5%, respectively. These parameters were used in the thermographic software to evaluate the air transmission coefficient.

The thermographic images were acquired at 6 Hz for the duration of the test. This frequency is suitable for the analysis, as respiratory and coughing phenomena evolve at speeds compatible with a 6 Hz sampling rate, ensuring that no relevant information is lost during acquisition. Then, a reference frame of the acquisition was subtracted from each image acquired. The subtraction allows to better highlight the thermal footprint since the temperature increase due to the saliva droplets is very low. In order to clearly individuate the reference frames, the mean temperature measured on the face mask was analyzed for each test. Fig. 2 shows the mean temperature measured for subject 1 wearing the surgical mask. It is worth noting that in this case, the first frame was taken as a reference for the subtraction. The points just before each temperature increase were associated with the breaths and coughs, as highlighted by the dashed lines in Fig. 2. Table 1 summarizes the times of breaths and coughs for each test. Hereinafter, the frame corresponding to the time of the first cough was taken as the reference frame for the image subtraction.

Finally, a window average filter, with an extension of 5 × 5 pixels, was applied to the images in order to make it easier to detect the thermal footprint and to perform a cleaning of the temperature data. Fig. 3 shows the comparison between the same frame with and without the filter application. It is easy to note that the filter makes the thermal footprint clearly detectable.

Two preliminary tests were performed in order to optimize the detection of the thermal footprint. First, the necessity to use the screen behind subjects was evaluated. The screen is made of a black fabric with emissivity equal to 0.9. Fig. 4 shows the thermal images captured with and without the screen. The thermal footprint was not detectable in the case without the screen; otherwise, the saliva droplet diffusion is visible when adopting the screen. It is worth noting that ambient temperature, distance of the subject from the screen, and the sequence of breaths and coughs were the same for the two images acquired.

The second test deals with the distance from the screen. Fig. 5 shows two infrared images with the subject placed at a distance from the screen of 10 cm (Fig. 5a) and 40 cm (Fig. 5b). The first detectable thermal footprint for both cases was represented in the figure. The thermal footprint is detectable for both of the two distances. However, in the case of higher distance from the screen, the thermal footprint has lower intensity. Furthermore, there is a higher delay between the cough and the thermal footprint detection. Indeed, Fig. 5a refers to the fourth frame after the first cough, which corresponds to a delay of about 0.5 s. Otherwise, Fig. 5b refers to the thirtieth frame after the first cough, which corresponds to a delay of about 4.8 s. For this reason it was decided to place the subjects at a distance of 10 cm from the screen, which allows a clear and easy detection of the thermal footprint. A shorter distance was considered inappropriate, as the subject’s presence could alter the temperature distribution on the screen or even result in accidental contact during a coughing event.

The analysis of experimental results was performed both from a qualitative and quantitative point of view. The infrared images allow to immediately detect the spread of the thermal footprint behind the subjects. From a first qualitative analysis of the infrared images, it was also possible to define the area interested by the saliva droplets, highlighted by the green lines in Fig. 6. The area was selected for the surgical mask tests and maintained the same for the tests with the FFP2 mask. In this way, it was possible to perform a quantitative comparison of the results. In particular, the time histories of maximum and mean temperatures inside these areas were calculated to analyze the experimental results. The maximum temperature value is related to the intensity of the thermal footprint, which corresponds to the amount of saliva droplet emission behind the subject: a higher maximum temperature means a more intense thermal footprint. The mean temperature deals with the spread of the thermal footprint: a higher mean temperature in the same area is related to a larger zone interested by the saliva droplets. Both parameters were employed to quantify the magnitude of leakage behind the subjects associated with mask use and, consequently, the potential infection risk to other individuals. Furthermore, the time interval between the cough event and the first peak in the time history of the maximum temperature difference was analyzed. This parameter provides an estimate of the droplet arrival time on the screen.

In this study, a novel non-invasive method was applied to investigate saliva droplet diffusion behind subjects during ordinary and extraordinary respiratory activities. The detection of the rearward leakage due to face masks is important to reduce the infection risk, especially for the emission in the head height zone. Specifically, the phenomenon was detected by measuring the thermal footprint left by saliva droplets on a screen positioned behind the subject. Infrared thermography was used to capture the thermal footprint. Furthermore, surgical and FFP2 protective face masks were tested to evaluate the influence of these two common types of masks on saliva droplet emission. The tests were performed with two subjects, one male and one female, who were asked to carry out a standard sequence consisting of two deep breaths and two coughs.

The results demonstrated the feasibility of the proposed method to detect the thermal footprint of saliva droplets and to address the leakage on the back of the subjects due to mask wearing. The behavior of particle emission was strongly influenced by the facial conformation of the subjects and the mask fit, as the shape and position of the thermal footprints differ significantly between the two individuals. The findings highlight that the surgical mask exhibits a greater saliva droplet emission and a larger affected area than the FFP2 mask. This is especially true in the head height area, while the FFP2 mask shows a leakage in the zone near the neck, which was not detected for the surgical mask case. It is important to remark that the nature of the phenomenon under investigation is inherently influenced by individual-specific characteristics. In this study the sample size is limited to two subjects, so the comparison between masks and subjects is only partially generalizable.

The thermography approach proposed in the paper could be a promising method to detect rearward leakage during ordinary and extraordinary respiratory activities, without any potential risk for the individual due to the non-invasive nature of the measurement technique.