A Low-Profile Ultrawideband Antenna Based on Flexible Graphite Films for On-Body Wearable Applications

This paper presents a low-profile ultrawideband antenna for on-body wearable applications. The proposed antenna is based on highly conductive flexible graphite films (FGF) and polyimide (PI) substrate, which exhibits good benefits such as flexibility, light weight and corrosion resistance compared with traditional materials. By introducing flaring ground and an arrow-shaped slot, better impedance matching is achieved. The wearable antenna achieves a bandwidth of 122% from 0.34 GHz to 1.4 GHz, with a reflection coefficient of less than −10 dB, while exhibiting an omnidirectional pattern in the horizontal plane. To validate the proposed design, the wearable antenna with a profile of ~0.1 mm was fabricated and measured. The measured results are in good agreement with simulated ones, which indicates a suitable candidate for on-body wearable devices.


Introduction
Wireless Body Area Networks (WBAN) have been widely applied in many aspects of novel wireless communication systems, e.g., health monitoring, military, and entertainment [1][2][3][4][5]. As a vital component in these systems, the wearable antenna is key to receiving the data from sensors while sending electric signals to the data register or base station for wireless wearable communication on or off-body channel communications (BCC) [6]. Ultrawideband technology is an attractive method for improved communications including wireless on-body networks. Various ultrawideband antennas have been reported such as Vivaldi form [7,8], microstrip form [9,10] and monopole form [11,12]. In these cases of single-layer antennas, it usually has a low profile, which shows great potential for wearable applications. Recently, wearable antennas have attracted much attention and have been rapidly developed [13][14][15][16][17]. However, there are several development bottlenecks for the design approach and performance requirements. The close vicinity of a lossy human body results in a reduction in antenna efficiency due to its high power consumption. Besides, the performance of a wearable antenna is sensitive to the different positions on the human body and different human bodies, such as men and women [18,19]. Simultaneously, the exposure of the human body to the electromagnetic field must be considered for safety. Therefore, a wearable antenna with a low specific absorption rate (SAR) is highly desired [20,21]. More importantly, for practical implementation, it is critical to design a flexible antenna that is bendable when worn by the user although the bending effects degrade the performance of antennas compared to their flat condition [22]. Prior research has reported lots of wearable button antennas with relatively small sizes, which can be easily mounted and make it unnecessary to fabricate using flexible materials [23][24][25]. However, flexible materials are in great demand when a wearable antenna operates at the lower frequencies. It can be noted that few of the proposed wearable antennas work at frequencies lower than 1 GHz where some of the frequency ranges are used for Medical Implant Communications Service

Characterization of Materials
In the field of wearable antenna design, various conductive materials are adopted to act as radiation structures such as conductive fabrics [9], copper [13], and textile-based materials [25,31], and so on. However, metallic materials exhibit unsatisfactory flexibility, while textile-based conductive materials show bad performance when exposed to humid circumstances. Compared with mentioned traditional conductive materials, the FGF shows excellent superiorities in light weight, favorable flexibility and corrosion resistance [32,33]. The FGF is prepared in the following three steps: vacuumed heating, firing under argon atmosphere, and rolling process, which is described in detail in [33]. The FGF we used in this paper comes from the Hubei Engineering Research Center of RF-Microwave Technology and Application in the Wuhan University of Technology. As shown in Figure 1, we describe the characteristic of the FGF in our previous work [32]. Figure 1a illustrates the cross-section scanning electron microscopy (SEM) image of the FGF sample, which shows the thickness of the FGF is~26 µm, and the interrelating electrical conductivity is 1.1 × 10 6 S/m, close to Materials 2021, 14,4526 3 of 14 the traditional metallic like copper (1.3 × 10 7 S/m). More importantly, the FGF is much lighter than the copper because the density of the FGF is 1.8 g/cm 3 , which is around a fifth of the copper density of 8.8 g/cm 3 [33]. The experiment of mechanical reliability of the FGF is demonstrated in Figure 1b. It can be seen that the FGF maintains its unchanged resistivity after 500 times bending, which proves the FGF has good flexibility and mechanical stability.
Technology and Application in the Wuhan University of Technology. As shown in Figure  1, we describe the characteristic of the FGF in our previous work [32]. Figure 1a illustrates the cross-section scanning electron microscopy (SEM) image of the FGF sample, which shows the thickness of the FGF is ~26 μm, and the interrelating electrical conductivity is 1.1 × 10 6 S/m, close to the traditional metallic like copper (1.3 × 10 7 S/m). More importantly, the FGF is much lighter than the copper because the density of the FGF is 1.8 g/cm 3 , which is around a fifth of the copper density of 8.8 g/cm 3 [33]. The experiment of mechanical reliability of the FGF is demonstrated in Figure 1b. It can be seen that the FGF maintains its unchanged resistivity after 500 times bending, which proves the FGF has good flexibility and mechanical stability. In addition, the textile fabric is a porous, anisotropic and compressible material, whose structures and electromagnetic properties can be affected severely by the surroundings [34], thus flexible polyimide (PI) substrate is preferred to textile material. In this work, we choose PI as the substrate with a thickness of 0.1 mm, a dielectric constant of 3.5 and a loss tangent of 0.0027 (purchased from DuPont company of America).

Antenna Design
Generally speaking, Microstrip antenna has a high quality factor (Q), which leads to a narrow bandwidth [35]. Compared with microstrip antenna, planar monopole antenna (PMA) is a common antenna that is widely utilized in wearable devices due to its simple structure, low profile and easy integration with the human body. The design progress of the proposed ultrawideband wearable antenna is exhibited in Figure 2. The initial antenna geometry is based on a typical planar monopole antenna. As depicted in Figure 2a, the prototype consists of a monopole patch and ground designed above the PI substrate. Both monopole patch and ground adopt the FGF for desired requirements. The proposed antenna is excited by a 50-ohm coplanar waveguide (CPW) for impedance matching, in which w and g are the widths of the central conductor and the gap between the central conductor and the ground, respectively. In addition, the textile fabric is a porous, anisotropic and compressible material, whose structures and electromagnetic properties can be affected severely by the surroundings [34], thus flexible polyimide (PI) substrate is preferred to textile material. In this work, we choose PI as the substrate with a thickness of 0.1 mm, a dielectric constant of 3.5 and a loss tangent of 0.0027 (purchased from DuPont company of America).

Antenna Design
Generally speaking, Microstrip antenna has a high quality factor (Q), which leads to a narrow bandwidth [35]. Compared with microstrip antenna, planar monopole antenna (PMA) is a common antenna that is widely utilized in wearable devices due to its simple structure, low profile and easy integration with the human body. The design progress of the proposed ultrawideband wearable antenna is exhibited in Figure 2. The initial antenna geometry is based on a typical planar monopole antenna. As depicted in Figure 2a, the prototype consists of a monopole patch and ground designed above the PI substrate. Both monopole patch and ground adopt the FGF for desired requirements. The proposed antenna is excited by a 50-ohm coplanar waveguide (CPW) for impedance matching, in which w and g are the widths of the central conductor and the gap between the central conductor and the ground, respectively. All the simulations of the antenna are based on the ANSYS high-frequency structure simulator (HFSS). The simulated reflection coefficient S11 of the proposed antenna is shown in Figure 3. The traditional monopole antenna (Ant-1) has realized a bandwidth of 109% from 0.34-1.16 GHz (i.e., S11 < −10 dB), in which two resonances are observed at 0.39 GHz and 0.97 GHz. It is worth mentioning that the original antenna has realized the ultrawideband feature. The distribution plots of surface current at resonant frequencies above are given in Figure 4. It can be seen that Ant-1 operates in half-wavelength mode at 0.39 GHz and full-wavelength mode at 0.97 GHz, respectively. The currents of both sides of the slot show opposite directions, indicating the current distribution similar to a slot antenna. Therefore, the antenna in this design can be equivalent to a monopole antenna combined with two slot antennas.  All the simulations of the antenna are based on the ANSYS high-frequency structure simulator (HFSS). The simulated reflection coefficient S 11 of the proposed antenna is shown in Figure 3. The traditional monopole antenna (Ant-1) has realized a bandwidth of 109% from 0.34-1.16 GHz (i.e., S 11 < −10 dB), in which two resonances are observed at 0.39 GHz and 0.97 GHz. It is worth mentioning that the original antenna has realized the ultrawideband feature. The distribution plots of surface current at resonant frequencies above are given in Figure 4. It can be seen that Ant-1 operates in half-wavelength mode at 0.39 GHz and full-wavelength mode at 0.97 GHz, respectively. The currents of both sides of the slot show opposite directions, indicating the current distribution similar to a slot antenna. Therefore, the antenna in this design can be equivalent to a monopole antenna combined with two slot antennas. All the simulations of the antenna are based on the ANSYS high-frequency structure simulator (HFSS). The simulated reflection coefficient S11 of the proposed antenna is shown in Figure 3. The traditional monopole antenna (Ant-1) has realized a bandwidth of 109% from 0.34-1.16 GHz (i.e., S11 < −10 dB), in which two resonances are observed at 0.39 GHz and 0.97 GHz. It is worth mentioning that the original antenna has realized the ultrawideband feature. The distribution plots of surface current at resonant frequencies above are given in Figure 4. It can be seen that Ant-1 operates in half-wavelength mode at 0.39 GHz and full-wavelength mode at 0.97 GHz, respectively. The currents of both sides of the slot show opposite directions, indicating the current distribution similar to a slot antenna. Therefore, the antenna in this design can be equivalent to a monopole antenna combined with two slot antennas.  To achieve a broader operating bandwidth, two rectangle slots are replaced with tapered slots by adopting flaring ground. As depicted in Figure 2b, the black dashed-dotted line corresponds to the outline of the flaring ground that satisfies the elliptical equation: where W1 and R1 (R2) represent the minor-axis and major-axis radii of the elliptical patch, respectively. It can be seen from Figure 3 that the modified monopole antenna (Ant-2) has a broader bandwidth above 1.16 GHz compared with Ant-1. However, we can also observe that the reflection coefficient does not reach −10 dB around 0.55 GHz, which leads to performance degradation of the antenna.
Here a solution is presented by inserting an arrow-shaped slot structure on the radiation patch. Note that the flaring ground and the arrow-shaped slot are adopted for the wideband impedance matching. The simulated surface current distribution of the proposed antenna (Ant-3) is plotted in Figure 5. The surface current of Ant-3 increases around the arrow-shaped slot at 0.97 GHz compared with the one obtained of Ant-1, which helps to reduce S11 for the fractional bandwidth from 109% to 122%. The variation of S11 with W2 is shown in Figure 6. The arrow-shaped slot generates a capacitance effect, which helps to improve the impedance bandwidth and realize antenna miniaturization. Moreover, the current path of the Ant-3 becomes longer when W2 increases. However, when W2 is too large, the capacitance effect becomes weaker, which provides few contributions to impedance matching. Finally, we select the optimal value of 15 mm. Eventually, we obtain a flexible ultrawideband monopole antenna. Figure 2c depicts the geometry of the final ultrawideband wearable antenna (Ant-3). As shown in Figure 3, a comparison with the S11 obtained from the proposed antenna exhibits that the operating bandwidth of Ant-3 has been effectively improved with the fractional bandwidth of 122%. The optimized geometrical parameters of the proposed antenna are listed in Table 1. To achieve a broader operating bandwidth, two rectangle slots are replaced with tapered slots by adopting flaring ground. As depicted in Figure 2b, the black dashed-dotted line corresponds to the outline of the flaring ground that satisfies the elliptical equation: where W 1 and R 1 (R 2 ) represent the minor-axis and major-axis radii of the elliptical patch, respectively. It can be seen from Figure 3 that the modified monopole antenna (Ant-2) has a broader bandwidth above 1.16 GHz compared with Ant-1. However, we can also observe that the reflection coefficient does not reach −10 dB around 0.55 GHz, which leads to performance degradation of the antenna.
Here a solution is presented by inserting an arrow-shaped slot structure on the radiation patch. Note that the flaring ground and the arrow-shaped slot are adopted for the wideband impedance matching. The simulated surface current distribution of the proposed antenna (Ant-3) is plotted in Figure 5. The surface current of Ant-3 increases around the arrow-shaped slot at 0.97 GHz compared with the one obtained of Ant-1, which helps to reduce S 11 for the fractional bandwidth from 109% to 122%. The variation of S 11 with W 2 is shown in Figure 6. The arrow-shaped slot generates a capacitance effect, which helps to improve the impedance bandwidth and realize antenna miniaturization. Moreover, the current path of the Ant-3 becomes longer when W 2 increases. However, when W 2 is too large, the capacitance effect becomes weaker, which provides few contributions to impedance matching. Finally, we select the optimal value of 15 mm. Eventually, we obtain a flexible ultrawideband monopole antenna. Figure 2c depicts the geometry of the final ultrawideband wearable antenna (Ant-3). As shown in Figure 3, a comparison with the S 11 obtained from the proposed antenna exhibits that the operating bandwidth of Ant-3 has been effectively improved with the fractional bandwidth of 122%. The optimized geometrical parameters of the proposed antenna are listed in Table 1.   There is no doubt that the bending effects of the wearable antenna degrade the performance compared with its flat condition, and the wearable antenna performance will be also affected by unavoidable deformation even though the antenna is designed for a certain bending radius. In this design, the proposed ultrawideband wearable antenna is conformally integrated over a cylindrical foam with a radius of R. By taking bending states of the body surface into account, we set the bending radii of R = 60 mm, 80 mm and 100 mm. As plotted in Figure 7, the cylindrical surface on which the antenna is bent is oriented   There is no doubt that the bending effects of the wearable antenna degrade the performance compared with its flat condition, and the wearable antenna performance will be also affected by unavoidable deformation even though the antenna is designed for a certain bending radius. In this design, the proposed ultrawideband wearable antenna is conformally integrated over a cylindrical foam with a radius of R. By taking bending states of the body surface into account, we set the bending radii of R = 60 mm, 80 mm and 100 mm. As plotted in Figure 7, the cylindrical surface on which the antenna is bent is oriented  There is no doubt that the bending effects of the wearable antenna degrade the performance compared with its flat condition, and the wearable antenna performance will be also affected by unavoidable deformation even though the antenna is designed for a certain bending radius. In this design, the proposed ultrawideband wearable antenna is conformally integrated over a cylindrical foam with a radius of R. By taking bending states of the body surface into account, we set the bending radii of R = 60 mm, 80 mm and 100 mm. As plotted in Figure 7, the cylindrical surface on which the antenna is bent is oriented along the y-axis. Figure 8 demonstrates the reflection coefficient of the wearable antenna at different R, which indicates a slight degradation of S 11 around 0.5 GHz and a shift of resonant frequency to lower frequency as well. It proves that the impedance matching of the wearable antenna is well maintained compared with its flat condition. along the y-axis. Figure 8 demonstrates the reflection coefficient of the wearable antenna at different R, which indicates a slight degradation of S11 around 0.5 GHz and a shift of resonant frequency to lower frequency as well. It proves that the impedance matching of the wearable antenna is well maintained compared with its flat condition.  To study the impact of the loading positions of the proposed antenna, the antenna performance in free space, thigh area and shank area is simulated. As depicted in Figure  9, the three-dimensional voxel model of the male of ANSYS HFSS is adopted to mimic real scenarios. The wearable antenna is located near the thigh area and shank area (10 mm above body surface) due to its dimension. This is done by attaching the antenna onto a cylindrical foam, with a diameter of 100 mm and 80 mm (not shown in the photograph), respectively. Figure 10 demonstrates the simulated reflection coefficient in free space, thigh area and shank area, which plots the resonant frequency shift to the lower frequency. Meanwhile, impedance mismatch occurs around 0.5 GHz when the antenna is mounted near the thigh area. Both of them are due to the high dielectric constant of the human body. along the y-axis. Figure 8 demonstrates the reflection coefficient of the wearable antenna at different R, which indicates a slight degradation of S11 around 0.5 GHz and a shift of resonant frequency to lower frequency as well. It proves that the impedance matching of the wearable antenna is well maintained compared with its flat condition.  To study the impact of the loading positions of the proposed antenna, the antenna performance in free space, thigh area and shank area is simulated. As depicted in Figure  9, the three-dimensional voxel model of the male of ANSYS HFSS is adopted to mimic real scenarios. The wearable antenna is located near the thigh area and shank area (10 mm above body surface) due to its dimension. This is done by attaching the antenna onto a cylindrical foam, with a diameter of 100 mm and 80 mm (not shown in the photograph), respectively. Figure 10 demonstrates the simulated reflection coefficient in free space, thigh area and shank area, which plots the resonant frequency shift to the lower frequency. Meanwhile, impedance mismatch occurs around 0.5 GHz when the antenna is mounted near the thigh area. Both of them are due to the high dielectric constant of the human body. To study the impact of the loading positions of the proposed antenna, the antenna performance in free space, thigh area and shank area is simulated. As depicted in Figure 9, the three-dimensional voxel model of the male of ANSYS HFSS is adopted to mimic real scenarios. The wearable antenna is located near the thigh area and shank area (10 mm above body surface) due to its dimension. This is done by attaching the antenna onto a cylindrical foam, with a diameter of 100 mm and 80 mm (not shown in the photograph), respectively. Figure 10 demonstrates the simulated reflection coefficient in free space, thigh area and shank area, which plots the resonant frequency shift to the lower frequency. Meanwhile, impedance mismatch occurs around 0.5 GHz when the antenna is mounted near the thigh area. Both of them are due to the high dielectric constant of the human body. along the y-axis. Figure 8 demonstrates the reflection coefficient of the wearable antenna at different R, which indicates a slight degradation of S11 around 0.5 GHz and a shift of resonant frequency to lower frequency as well. It proves that the impedance matching of the wearable antenna is well maintained compared with its flat condition.  To study the impact of the loading positions of the proposed antenna, the antenna performance in free space, thigh area and shank area is simulated. As depicted in Figure  9, the three-dimensional voxel model of the male of ANSYS HFSS is adopted to mimic real scenarios. The wearable antenna is located near the thigh area and shank area (10 mm above body surface) due to its dimension. This is done by attaching the antenna onto a cylindrical foam, with a diameter of 100 mm and 80 mm (not shown in the photograph), respectively. Figure 10 demonstrates the simulated reflection coefficient in free space, thigh area and shank area, which plots the resonant frequency shift to the lower frequency. Meanwhile, impedance mismatch occurs around 0.5 GHz when the antenna is mounted near the thigh area. Both of them are due to the high dielectric constant of the human body.  The simulated realized gain and radiation efficiency are illustrated in Figure 11a,b, respectively. We can observe that the realized gain of the antenna is not decreased severely due to the electromagnetic reflection of the body. However, as shown in Figure  11b, the radiation efficiency of the wearable antenna reduces from 95% to 60% because of the proximity of the lossy human body, which limits the power transmission in free space. As displayed in Figure 12, the wearable antenna has an omnidirectional pattern in the xoz plane and 8-shape patterns in the yoz plane when loaded near the human body. Besides, the gain at 0° direction is higher than the gain at 180° direction, which is mainly caused by the absorption of the human body. The simulated realized gain and radiation efficiency are illustrated in Figure 11a,b, respectively. We can observe that the realized gain of the antenna is not decreased severely due to the electromagnetic reflection of the body. However, as shown in Figure 11b, the radiation efficiency of the wearable antenna reduces from 95% to 60% because of the proximity of the lossy human body, which limits the power transmission in free space. As displayed in Figure 12, the wearable antenna has an omnidirectional pattern in the xoz plane and 8-shape patterns in the yoz plane when loaded near the human body. Besides, the gain at 0 • direction is higher than the gain at 180 • direction, which is mainly caused by the absorption of the human body.  The simulated realized gain and radiation efficiency are illustrated in Figure 11a,b, respectively. We can observe that the realized gain of the antenna is not decreased severely due to the electromagnetic reflection of the body. However, as shown in Figure  11b, the radiation efficiency of the wearable antenna reduces from 95% to 60% because of the proximity of the lossy human body, which limits the power transmission in free space. As displayed in Figure 12, the wearable antenna has an omnidirectional pattern in the xoz plane and 8-shape patterns in the yoz plane when loaded near the human body. Besides, the gain at 0° direction is higher than the gain at 180° direction, which is mainly caused by the absorption of the human body.

Measurement and Results
As shown in Figure 13, one prototype of the wearable antenna was fabricated and measured to validate the proposed antenna design. A LPKF laser engraver machine is used to outline the FGF-based monopole patch with desirable shapes on a PI substrate, which is described in detail in [33]. The wearable antenna and standard gain log periodic antenna were connected to the vector network analyzer (VNA). As presented in Figure  14, the bending effect to the antenna is negligible, which not only exhibits the great property of the material such as flexibility and mechanical stability, but also indicates the stable performance of the antenna under bending conditions. To prove the practicability of the wearable antenna, the fabricated antenna was attached to two positions including the thigh area and shank area. Figure 15 provides the measured reflection coefficient, which shows a good agreement with the simulated result, except for a slight frequency deviation in the high-frequency band. It may be caused by manufacturing errors and different dielectric constants corresponding to different positions. It is worth noting that the real human body increases the impedance matching bandwidth in the low-frequency band compared with numerical results, which indicates that most of the input power has been transmitted to the antenna. The simulated and measured radiation patterns at 0.4 GHz and 1.2 GHz are given in Figure 16, which has exhibited an omnidirectional pattern in the xoz plane as well as reached a good consistency, despite slight degradations in minor direc-

Measurement and Results
As shown in Figure 13, one prototype of the wearable antenna was fabricated and measured to validate the proposed antenna design. A LPKF laser engraver machine is used to outline the FGF-based monopole patch with desirable shapes on a PI substrate, which is described in detail in [33]. The wearable antenna and standard gain log periodic antenna were connected to the vector network analyzer (VNA). As presented in Figure 14, the bending effect to the antenna is negligible, which not only exhibits the great property of the material such as flexibility and mechanical stability, but also indicates the stable performance of the antenna under bending conditions. To prove the practicability of the wearable antenna, the fabricated antenna was attached to two positions including the thigh area and shank area. Figure 15 provides the measured reflection coefficient, which shows a good agreement with the simulated result, except for a slight frequency deviation in the high-frequency band. It may be caused by manufacturing errors and different dielectric constants corresponding to different positions. It is worth noting that the real human body increases the impedance matching bandwidth in the low-frequency band compared with numerical results, which indicates that most of the input power has been transmitted to the antenna. The simulated and measured radiation patterns at 0.4 GHz and 1.2 GHz are given in Figure 16, which has exhibited an omnidirectional pattern in the xoz plane as well as reached a good consistency, despite slight degradations in minor directions. Table 2 lists the comparison between the presented wearable antenna and other existing wide-band wearable antennas. It can be concluded that our work realizes ultrawideband characteristics in the sub-GHz region with the lowest profile. More importantly, the adoption of the FGF and PI substrate guarantees the favorable flexibility, mechanical stability and light weight of the wearable antenna. tions. Table 2 lists the comparison between the presented wearable antenna and other existing wide-band wearable antennas. It can be concluded that our work realizes ultrawideband characteristics in the sub-GHz region with the lowest profile. More importantly, the adoption of the FGF and PI substrate guarantees the favorable flexibility, mechanical stability and light weight of the wearable antenna.
(a)  tions. Table 2 lists the comparison between the presented wearable antenna and other existing wide-band wearable antennas. It can be concluded that our work realizes ultrawideband characteristics in the sub-GHz region with the lowest profile. More importantly, the adoption of the FGF and PI substrate guarantees the favorable flexibility, mechanical stability and light weight of the wearable antenna.         Where λ 0 is the free space wavelength corresponding to the center frequency.

Conclusions
A novel ultrawideband antenna using highly conductive flexible graphite films (FGF) for on-body wearable applications is proposed, fabricated and tested. The ultrawideband characteristic is realized by modifying the planar monopole antenna using the flaring ground and the arrow-shaped slot. In addition, to satisfy some special performance of the wearable antenna, we adopt the FGF as the conductive material, which shows a good candidate for wearable devices. To the best of our knowledge, it is the first time to adopt the FGF to replace traditional metallic materials in wearable antenna design within the sub-GHz region. The fabricated prototype has only a profile of 0.1 mm, making it applicable for conformal and wearable applications. The measured results have demonstrated that the FGF antenna operates from 0.34 GHz to 1.4 GHz with fractional bandwidth of 122% while providing an omnidirectional vertical polarized pattern in the xoz plane although it is bent and working in the proximity of the human body. The proposed antenna has many superiorities such as low profile, favorable flexibility, light weight, ultrawideband, and omnidirectional beam pattern, which may be suitable for an on-body communication system. This work has also demonstrated that the FGF has great potential in wearable antenna design. Institutional Review Board Statement: Ethical review and approval were waived for this study since the wearable antenna was worn on the surface of clothes and excited by minimal power (<−3 dBm), which shows negligible impact on the human body according to the IEEE C95.1-1991 [21]. Moreover, the wearable antenna obtained the informed consent of all the researchers and subjects in this project, so we did not need ethical review and approval for this study.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. Data Availability Statement: Not applicable.