The real-time TMCF monitoring system basically consists of four force sensor units, a microcontroller board (Arduino Bluno Mega 2560), a control PC, and an in-house software program (Figure 1). In this study, a SingleTact™ capacitive force sensor manufactured by Pressure Profile System, Inc. (Glasgow, UK) was used. SingleTact™ is a capacitive force sensor with a parallel capacitive configuration, and it is considered to be suitable for this application due to its high accuracy and repeatability (Tang et al., 2020). SingleTact™ consists of two thin round polyimide electrode plates separated by a dielectric. The working principle of the sensor is based on flexible interlayer compression as a result of the force that arises from distance change between two electrodes causing capacitance variation (Graudone et al., 2020). SingleTact™ includes electronics providing an interface to the main controller. The supply voltage needs to be between 3.7 V and 12 V with an input current of 2.7 mA. The range of the analog output voltage is between 0.5 V and 1.5 V. In terms of data transfer, SingleTact™ can accommodate more than 100 Hz (Warsito et al., 2020). For real-time TMCF monitoring, a calibrated sensor CS8-10N with a diameter of 8 mm was used. Calibrated sensors offer improved accuracy and linearity over standard sensors and come as a matched sensor plus electronics interface board providing a pre-configured system in terms of the electronic interface (Figure 2.) and with calibration carried out for linear output. Pre-calibrated sensors can measure pressure up to over 1500 gf/cm² with a resolution of about 0.2% of the full scale. The response time of the system is 1 ms (Pressure Prole Syst., Inc, 2022a, User Manual).

While the thickness of the sensor itself is 0.35 mm, two sponge pads of about 3 mm (that is, 1.5 mm each), called pressure absorption layer (PAL), are added, one above and the other below, to improve contact sensitivity between the mask and patient skin as shown in Figure 2. A thermoplastic mask, in an ideal situation, should be formed tightly to the patient skin without gaps. In reality, however, there are often small gaps (about 1 mm). Therefore, an appropriate thickness of PAL can help improve the contact sensitivity between the mask and the patient skin. The PAL thickness needs to be adjusted depending on the position to be monitored and the degree of contact between the mask and the face. It is also instrumental to select an appropriate material for PAL.

Initially, we tried to use silicon material (Dragon Skin TM 30) with stiffness similar to that of skin (30 Shore A). However, its stiffness turned out too strong to be used for mask force monitoring (the sensor signal got saturated easily). Therefore, the PAL material was changed to a soft sponge (about 60 Shore OO, stiffness similar to that of sliced bread). In order to find an appropriate PAL thickness that can reliably measure the mask force within the active range of the sensor, the thickness of the sponge was increased by a 1-mm interval.

When the PAL is too thin (1 mm), as compared to the optimal value (3 mm), movements inside the mask would not be monitored effectively. When the PAL was 1 mm, TMCF baseline values were observed between 80 and 120 gf/cm². As can be seen in Figure 5, this value is the force when the soft sponge (PAL) is compressed by about 0.1 mm. Considering the stiffness of the soft sponge, it can be seen that the value is an insufficient force to keep the TMCF stable.

Actually, in real-time monitoring conducted on six volunteers, it was confirmed that the lower the baseline of TMCF, the more difficult it is to maintain the compression force stable. In addition, it has been confirmed that if the TMCF falls below 100 gf/cm² (that is, when TMCF is too low), the compression force can no longer be maintained and often drops to zero. In other words, 100 gf/cm² seemed a threshold value when the mask was about to lose meaningful contact with the surface of the volunteer. Therefore, the value of 100 gf/cm² was applied as the first tolerance. The current version of GUI displays a yellow light warning to alert the user when the real-time TMCF falls below 100 gf/cm² (first tolerance). Also, if TMCF drops to ‘0' (2nd tolerance), it is designed to give a red-light warning.

In this study, every mask was made by a single experienced radiation therapy technologist to keep consistency and about 3 mm of PAL was found appropriate in effectively monitoring mask force. When the PAL is too thick (over 5 mm PAL), as compared to the optimal value (3 mm), the sensor would encounter frequent signal saturation. Thus, the total thickness of the force sensor unit (sensor and the optimal value of PAL) is about 3.35 mm (0.35 + 3 mm). Note, in actual practice, the optimal PAL thickness may vary depending on what type of mask is used and how it is formed.

The sensor units were placed between the thermoplastic mask and the patient face at four sensing points (forehead, both cheeks, and chin). It is recommended to avoid bending a sensor if possible because when bent, one electrode would be in compression while the other in tension, which can cause shearing to occur between the two electrodes. When bending is unavoidable, it is recommended to keep the circular portion (the active sensing region) as flat as possible while applying the bend along the tail end of the sensor. It is also necessary to keep a bend radius greater than 3 mm (Pressure Prole Syst., Inc, 2022b, SingleTact Mounting Methods). Therefore, caution is needed when placing sensors, especially on the curved parts of the patient face.

The analog signal produced by the capacitive sensor is digitized with an Arduino Bluno Mega 2560 and connected to a PC for data processing and analysis (Figure 1). The in-house developed software program, written in LabView, continuously monitors compression forces at four points and performs data processing to indicate whether the pressure at each point is within tolerance or not.

In this test, we tried to evaluate whether the same pressure values were obtained repeatedly when the sensor unit was pressed with known forces. A micrometer with a resolution of 0.001 mm was used to compress the PAL consistently. Figure 3 shows the experimental setting for the sensor unit reproducibility test. The procedure is as follows:

1) Place a piece of mask and the sensor unit between the anvil and spindle of the micrometer (Figures 3A,B).

Sensor output was recorded every 100 milliseconds for 1 min in each trial, and a total of 10 trials were made for all four sensor units.

Constant compression was given for 20 min (a time interval comparable to a typical radiation therapy session), and pressure change with time was measured.

In addition, a reproducibility test was performed to evaluate the reliability of the developed monitoring system prior to a volunteer test. Under an ideal condition, compression force inside the mask is expected not to vary in multiple uses. To closely simulate an ideal situation, a stability test was performed with a rigid head phantom. Figure 4 shows the experimental setting for the system reproducibility test. Through the test, it was tried to confirm whether the same compression force was obtained repeatedly when the rigid head phantom was immobilized with the system in multiple trials. In each trial, the data were acquired every 100 milliseconds for 3 min (that is, 1800 samples for each trial), and then the mean value was recorded. This process was repeated for a total of five times, and the mean SD value was obtained.

A total of six healthy volunteers (five males and one female) were enrolled in this simulation test. The volunteer group included two medical physicists, three graduate students, and one from the general public. The age ranged from 26 to 41 yrs.

For each volunteer, a thermoplastic mask was formed, and four force sensor units were attached to the protrusions of the face (that is, forehead, both cheekbones, and chin) using a transparent adhesive film (Tegaderm Film, 3M). After completing the mask setup, compression force values were monitored for 3 min in real time, which was considered one session. Each volunteer took four sessions, and, in between, the mask was off and an interval of approximately 10 min was given.

To investigate the effectiveness of the system for detecting abrupt force/motion, volunteers were instructed to try to move intentionally to simulate situations where a relatively large force of motion and/or variation occurred within the mask. Attempts of intentional movement (for example, nod motion with short bursts of force) were made at 3 specific times (100, 300, and 500 msec after monitoring started). Compression force values were observed for 1 min and compared with the results of the previous test (that is, Section 2.4).

Table 1 shows pressure values according to the compressed distance of the force sensor unit. In this test, we evaluated how consistent the monitored pressure values were among 10 trials. While absolute SD values tended to increase as the units were compressed more, relative ones decreased down to about 2.4% (that is, about 36 gf/cm²) of 1500 gf/cm² (the max of full scale). Overall, less than about 5% of SD was observed with pressure values larger than approximately 300 gf/cm². It was also found that pressure reading got almost saturated when the units were compressed to 1.1 mm or more.

Mean values in Table 1 are plotted according to compression distances in Figure 5. As it is shown in the figure, it could be observed that the differences between the measured forces for corresponding compressions increase with the compression since the stiffness of the PAL sponge increases as it is compressed (even if the PAL was compressed evenly to 1 mm, the pressure applied to the sensor gradually increased). However, as illustrated, a clear linearity between the obtained pressure and applied compression was obtained (R² value over 0.98 obtained for all sensors).

The pressure drift of the SingleTact sensor fluctuates between the zero points, thus giving positive and negative results intermittently. No obvious trend of the pressure drift can be observed (Figure 4).

In the time effect test with the head phantom, since there was neither contour change nor motion, constant compression forces of the mask were expected, and the system showed stable readings as shown in Figure 6.

Table 2 shows the results of the mask setup reproducibility test in which a total of five 3-min trials were repeated using the rigid head phantom. Because setups were attempted with the rigid object, if a precise mask setup was performed, compression force variation should be minimal. As summarized in Table 2, the standard deviation of each measuring spot did not exceed 39 gf/cm² (2.6% of 1500 gf/cm²).

Figure 7 shows the compression force variation of each volunteer (result of the first session). Volunteers were pressed using a customized mask, and then the compression force variation was monitored for 3 min per session for four sessions. As can be seen in the figure, it was observed that the compression force varied in a noticeable amount with time in general unlike the cases of the previous test with the head phantom (Figure 6). The compression force of the forehead was generally higher and more stable than that of the other parts in most volunteers. Compression force values for both cheekbones were also generally stable in all volunteers. However, in the case of the chin, the overall compression force was lower than that in other regions. In addition, its variation was more dramatic and sometimes reached down to zero (0).

Figure 8 shows the variation range and mean value of compression force for each session. Volunteers 1, 2, 4, and 6 maintained high compression forces (>800 gf/cm²) at all sessions, and the compression forces of volunteers 3 and 5 also did not drop below 600 gf/cm². In the case of both cheekbones, the variation ranges were slightly larger than those of the forehead case, but all volunteers maintained high compression forces in general (over 500 gf/cm²) in all sessions. In the chin case, the mean values were lower than those of other sites, and the ranges were noticeably larger in general. However, variations among sessions were not significant.

Figure 9 shows compression force variation due to intended motion inside the thermoplastic mask. As shown, noticeable changes in the compression force were observed at times when movement was instructed (that is, at 100, 300, and 500 ms), indicating the capability of the system to monitor possible patient motion and/or displacement inside the mask.

As mentioned previously, the GUI of the monitoring system is designed to give a yellow light warning signal when the real-time compression force value falls below 100 gf/cm². Also, if the compression force value drops to “0”, it is designed to give a red-light warning signal. Therefore, if the yellow light is on, the patient should be observed more carefully, and if the red light is on, a readjustment is recommended because it can be assumed that the contact between the mask and face is not appropriate.