An Implantable Wideband Microstrip Patch Antenna Based on High-Loss Property of Human Tissue

In this paper, an approach was designed to develop an implantable wideband microstrip patch antenna (MPA) by the high-loss property of human tissue. In order to reduce the detuning effect, an extensive investigation was conducted, demonstrating the effects of high-loss human tissue on the total quality factor (<inline-formula> <tex-math notation="LaTeX">$Q_{T}$ </tex-math></inline-formula>), effective relative permittivity, and effective loss tangent of the deeply implanted MPA. To the best of our knowledge, the exact three parameters for the deeply implanted MPA are firstly studied. From the evolutionary point of view, a deeply implanted MPA can be treated as a resonator with the extremely low and stable <inline-formula> <tex-math notation="LaTeX">$Q_{T}$ </tex-math></inline-formula>. As such, its wide impedance bandwidth can be easily achieved through two inherent radiative modes of the embedded MPA under the high-loss environment. Finally, the proposed MPA will be fabricated and tested, and the final outcomes are consistent with the corresponding simulated results.


I. INTRODUCTION
Recently, there is an increasing interest in designing and developing the wirelessly linked implantable medical devices (IMDs) for its applications in diagnosis and treatment towards effective communication with in-vitro monitors [1], [2]. An implantable antenna is generally taken as a key component of the biotelemetry system. Many kinds of the designed antennas can be embedded into human body, such as dipole [3]- [8], loop [9]- [11], helix [12], [13], microstrip patch antenna (MPA) [14]- [26], etc. Compared with other antennas, MPA is highly preferred for its easy fabrication, low profile, and integration with other electronic devices [14], [15].
The essential requirement for an implantable antenna is its stability of the resonant performance in human tissue. The performance of an implantable antenna strongly depends on the individuals' anatomical features of the surrounding The associate editor coordinating the review of this manuscript and approving it for publication was Cihun-Siyong Gong . tissues [7], [27]. Meanwhile, the dielectric properties of human tissue may be changed with multiple factors, such as gender, age, weight [7], etc. Thus, the detuning effect was identified as a great challenge for the design of an implantable antenna [28]. In general, a broadband impedance matching is highly demanded to compensate for an uncontrollable frequency shift [26]- [28].
So far, several research groups have made great significant contributions to design works of the wideband implantable MPAs. First of all, the impedance bandwidth of MPA is improved by the increased thickness of the substrate [29]. However, the intrinsic low-profile of MPA would be destroyed, resulting in the discomforts of patients. Besides, there are disadvantages of the limited bandwidth and the increased height of the antenna. Secondly, the additional resonant modes are proven to be effective. A variety of slots etched on the patch and ground [15]- [19], parasitic and stacked patches [20]- [26] are used to excite the non-inherent modes under its dominant radiative mode [30]. Reports show that the widest impedance bandwidth of an implantable MPA may reach up to about 47% through etching an openend slot on the ground [17]. However, the radiation patterns of the additional modes may be substantially changed with frequency and suffer from high cross-polarization [30]. To sum up, most of the published works are focused on broadening the impedance bandwidth of the antennas, but any insight into human tissue has not been comprehensively taken. Previously, the implantable MPAs with broadband performance are designed for shallow implantation, for instance, less than 5 mm of thickness, and low-frequency band (i.e., 400 MHz). In this paper, a design methodology based on the high-loss property of human tissues is firstly proposed to cope with the detuning effect for the deeply implanted MPA with the resonance of 2.4 GHz.
In Section II, the quantitative analysis is used to study the impaction of the high loss radiation environment on MPA. Except for effective relative permittivity (ε er ) and effective loss tangent (tanδ e ), the total quality factor (Q T ) of the deeply implanted MPA can be determined. Three important factors of implantation depth are investigated, demonstrating the stability of the embedded MPA in human tissue. To the best of our knowledge, the basic design theory about the deeply implanted MPA is firstly introduced to broaden the impedance bandwidth. In Section III, a differential-fed MPA with its distinctive Q T and input impedance is designed to monitor the pacemaker through the high-loss property of human tissue. In addition, the dual-band radiated modes are created for an implantable MPA with wide bandwidth. Sensitivity analysis was further carried out in terms of the different implantation depth, conductivity, and relative permittivity. Then the model integrity is studied to evaluate the practicality of the proposed MPA. In Section IV, the examination results of the fabricated antenna are accessible for the simulations. The conclusion can be eventually drawn in Section V.

II. PROPERTIES OF HUMAN TISSUE FOR IMPLANTABLE MPA
The goal of our work was to quantify the impaction of human tissue on the embedded MPA and increase its functional performances. To be specific, the total quality factor (Q T ), effective relative permittivity (ε r ), and effective loss tangent (tanδ e ) of MPA in human tissue would be analyzed. First and foremost, the implantable antenna played a key role in the wireless communication between IMDs and in-vitro monitors. Thus, the implantable MPA and practical IMD should be jointly designed [31]- [34]. Herein, the metal case of the pacemaker was taken as the ground plane of the proposed MPA [31], [34]. Fig.1 illustrates the entire geometry of a pacemaker with the addition of an implantable MPA. The size of the pacemaker was 40 mm × 40 mm ×10 mm, as it was referred from [33], [34].
Next, the embedded environment of IMDs might produce a significant effect on the performance of the implantable antenna, which was totally different from free-space. The radiated region of an in-body antenna referred to human tissue of which the heterogeneous media consisted of multiple  layers of tissue with frequency-dependent dielectric properties [27], [28]. In order to quantitatively analyze the influence of human tissue on the antenna, a useful model based on a single-layered tissue was utilized [1], [2]. Consequently, a frequency-dependent one-layered skin model [14] could be obtained, as it is shown in Fig. 2. The pacemaker with an implantable MPA was embedded in the center of this skin model. The distance from the structural side to the edge of the skin model was 30 mm at both sides. The distance from MPA to the bottom of the skin model was 20 mm. Then our proposed MPA was embedded in this tissue model with a depth of h s ; meanwhile, its upper surface was oriented towards the skin surface. The size of the cubic skin model was set as 100 mm × 100 mm × (20 + h s ) mm.

A. Q T OF IMPLANTABLE MPA
As it was mentioned above, the increase of impedance bandwidth was an effective approach to antagonize the detuning effect. Since Q T was inversely proportional to impedance bandwidth, the relationship between Q T and impedance bandwidth (BW) could be expressed as follows [29]:  where VSWR is the voltage standing-wave ratio and usually 2:1. Thus, it is figured out that Q T is the critical factor of broadening the impedance bandwidth of MPA. Generally, Q T could be reduced by thickening the substrate, etching slots on the patch and ground. However, more attention was paid to the design of an antenna itself. Few works were reported to make quantitative analysis about the influence of the surrounding environment on the implantable MPA. In this study, these influence factors should be elaborated in the design of the implantable MPA.
The physical model of the implantable MPA is illustrated in Fig. 3. In this model, the case of the pacemaker was acted as the reflector with the thickness of 0 mm. Through the simplified calculation process, our attention could be paid to study the influence of human tissue on the performance  of the implantable MPA. In fact, the size of the case was electrically large enough to suppress the edge effect of MPA. As a result, the thickness of the ground produced little effect on the radiation performance [35]. It was supposed that an infinitely thin ground was used. The simplified model was composed of a traditional MPA for being embedded into the skin. The MPA itself was printed on Rogers RO6010 with the thickness of h sub = 0.635 mm and the resonance of 2.4 GHz. Unless otherwise specified, all of the proposed MPAs used the same thickness of the substrate. In Fig. 3, l p is the physical length of the non-radiation edge while the length of the radiation edge is only one half of l p . MPA was directly fed by a 50 coaxial probe near the radiated edge. d f was the distance between the center of the coaxial probe and the edge, i.e., 0.5 mm.
According to [6] and [35]- [37], Q T could be obtained by the following formula: where ω 0 is the radian frequency; ω 0 = 2πf ; f is the working frequency. R in (ω 0 ) and X in (ω 0 ) are the input resistance and reactance of MPA; R in (ω 0 ) and X in (ω 0 ) are the corresponding slopes. Thus, Q T could be expressed by the input impedance of MPA; in addition, the required four parameters could be simulated by HFSS.

B. DISCUSSION OF THE RESULTS
The required parameters on the right side of Eq. (2) are shown in Fig. 4 and Fig. 5, respectively. In this paper, the calculation process was divided into two stages: shallow implantation (h s ≤ 5 mm), and deep implantation (h s > 5 mm). Fig. 4(a) and Fig. 4(b) illustrate the calculated input impedance of the shallowly implanted MPA and those of the deeply implanted MPA in Fig. 5. The input impedance is dramatically reduced with the increasing implantation depth (h s ) in Fig. 4, which are almost the same in Fig. 5. While substituting the required values in Eq. (2), the desired Q T with different implantation depths could be obtained in Fig. 6. In order to ensure the consistency of the skin with frequency-dependent dielectric properties, f was readily fixed at 2.4 GHz. As it could be seen in Fig. 6, Q T was substantially decreased with the depth h s of MPA in the rapidly varied region (shallow implantation, h s ≤ 5 mm). In the stable region (deep implantation, h s > 5 mm), Q T was almost identical. Results indicated that the deeply implanted MPA was actually a lossy resonator with extremely low and stable Q T , which might be totally different from those of other conventional MPAs in the air with high Q T . In general, the extremely low and stable Q T was generated by the huge loss of the loaded skin. As it is shown in Fig. 7, a standard MPA is produced with loading of a high loss superstrate when MPA is embedded into human tissue. Q T could be calculated by four types of losses [35] as follows: It was acknowledged that the surface-wave loss Q s could be generally neglected in the thin substrate; the conductor loss Q c was a constant; the radiation loss Q r was determined by the thickness and ε er ; the dielectric loss Q d was inversely proportional to the loss tangent tanδ e . The detailed calculation process for ε er , tanδ e , and Q T in the shallowly implanted MPA (h s ≤ 4 mm) were reported in [35]. The influence of human tissue on the deeply implanted MPA (h s > 5 mm) had never been reported before. Therefore, a quantitative study about the deeply implanted MPA was performed in this paper. Fig. 8 indicates the obtained values of ε er and tanδ e . Similarly, ε er and tanδ e were stable in the deeply implanted MPA. In contrast to tanδ e in the air, the corresponding value in the skin was sharply increased. The dominant factor Q T was changed from Q r to Q d , and related values were determined by tanδ e . In this paper, the high-loss properties of human tissue were primarily introduced for the design of an implantable wideband MPA.
The stable input impedance, ε er , tanδ e , and Q T in deep implantation brought an interesting point of view in the design of implantable antennas. According to Eq. (1), it can be well understood that Q T is opposite to impedance bandwidth. From Fig. 6, Q T is dramatically reduced with the increasing implantation depth in the shallowly implanted MPA. Thus, to increase the implantation depth is an easy approach to broaden the impedance bandwidth of the implantable MPA (h s ≤ 5 mm). Due to the stable Q T , there was an intrinsic limitation for deep implantation, that was, how to further broaden the impedance bandwidth. In this context, the high-loss property with the extremely low and stable Q T allowed the slow variation of the input impedance of the implantable MPA in human tissue within the desired operating band so that the inherent radiated modes could be facilely excited. Therefore, the so-called detuning effect might be effectively solved under the wideband performance of the implantable MPA.
Typically, the pacemaker was implanted with a depth of about 25.4 mm under the skin surface [2] so that the abovementioned characteristic could be stably maintained. However, the implantation depth of IMD in millimeters could hardly be controlled in current medical conditions. Thus, the tolerance to the variation of implantation depth should be further investigated for the implantable MPA in human tissue. As it was mentioned above, the deeply implanted MAP could be developed in a similar way.

III. ANTENNA DESIGN AND PARAMETRIC STUDY
Our next target is to design an implantable MPA in the deep implantation with wide impedance bandwidth that overcomes the detuning effect by means of the inherent high loss properties of human tissue. The proposed MPA can be operated at the 2.4-2.48 GHz industrial, scientific, and medical (ISM) band. Fig. 9 shows the geometrical schematic of the proposed differential-fed MPA with the detailed parameters, as denoted in Table 1. For convenience of application, the origin of the coordinate system serves as the center of the implantable MPA. It can be seen from Fig. 9 that a pair of 50 coaxial probe with equal amplitude and 180 • phase difference [30], [38] is installed to feed this MPA with a pair of onequarter-wavelength impedance transformer. The proposed MPA is implanted in the one-layered skin model with 20 mm VOLUME 8, 2020   Fig. 9. embedded depth, as indicated in Fig. 2. Here, the used substrate is still Rogers RO6010 with a thickness of 0.635 mm.

A. OPERATING PRINCIPLE
In order to communicate with in-vitro monitor, the radiation peak of the proposed MPA should be kept at boresight. The differential signals can effectively suppress those even-order modes, e.g., TM 20 and TM 22 , with null in broadside direction. Meanwhile, the length of the edge was adjusted in order to remove the TM 12 mode out of the operating band. The process of removing or suppressing those undesired modes has been extensively reported in [30]. As a sequence, only TM 10 and TM 30 modes were resonated in the operation range.
According to the discussion in Section II, the input impedance of the deeply implanted MPA had a very small value and a smooth variation, making this property useful for impedance matching. Then, a pair of microstrip line was introduced to feed this MPA. Fig. 10 illustrates the simulated R in of a differential-driven implanted MPA with different lengths of microstrip line (l i ). From Fig. 10, it can be seen that the resonant frequency of TM 30 mode (f 30 ) tends to decrease with the lengthening l i , and the corresponding value of TM 10 mode (f 10 ) remains approximately constant. In Fig. 11, the radiated patch with the reflected ground is connected by the two pairs of short pin in the diameter of 0.5 mm. By widening the distance between the two pairs of short pin (x s ), f 10 was increased, and the R in between the TM 10 and TM 30 modes was raised. In such a way, good impedance matching around the designed frequency can be achieved by the input impedance of the proposed MPA.

B. SIMULATED RESULTS
The simulated reflection coefficient (|S dd11 |) [30] of the differential-driven implantable MPA is shown in Fig. 12. In Fig. 12, there is the emergence of the two distant radiated modes, i.e., TM 10 and TM 30 modes. To the greatest extent, the simulated |S dd11 | bandwidth below -10 dB was ranged from 2.09 and 4.14 GHz (65.8% infraction). The broad bandwidth was achieved by means of the tissue's impaction on deeply implanted MPA. Fig. 13 and Fig. 14 show the realized gain   patterns at two resonant modes, including 2.41 and 3.61 GHz. The proposed MPA had a peak gain of −17.3 dBi and was radiated in the Z -axis direction. The high loss of the surrounding human tissue made contributions to the negative gain. With four planes, the co-polarizations were symmetrical and the cross-polarization suppression was suppressed lower than -46 dBi, respectively.
As for designing the implantable equipment, people's safety should be the largest issue. To realize such a goal, the IEEE standard of C95.1-1999 and C95.1-2005 not only assessed the RF radiation safety but also set a restriction on the specific absorption rate (SAR) levels [20]. In other words, two standards should be observed by all the designed implantable antenna. The standard of C95.1-1999 restricts  that the SAR levels averaged over 1-g of human tissue should be less than 1.6 W/kg, while the standard of C95.1-2005 indicates that the SAR levels averaged over 10-g of tissue should be less than 2 W/kg. The former standard was applied in this work since the standard of C95.1-1999 is much tougher than that of the C95. . The simulated value of maximum averaged SAR in 1-g of human skin at 2.4 GHz was presented by the first column of Table 2, which will be 226.5 W/kg if the input power was 1 W. To observe with the standard of IEEE standards of C95.1-1999, the maximum input power of the proposed MPA should be less than 9.5 mW in one-layered skin model, as it was shown in the third column of Table 2.

C. SENSITIVITY ANALYSIS
As it was mentioned above, an implantable antenna was always embedded in human tissue, which makes its resonant performance primarily dependent on the surrounding tissue. As a result, the antenna's resonance under various tissueloading conditions should be investigated in order to make a right evaluation, i.e., relative permittivity (ε r ), implant depth (h s ), and conductivity (σ ).
On the one hand, there are different dielectric properties of human tissue, enabling the performances of implantable antennas to be largely influenced by tissue variations. To quantitatively evaluate the proposed MPA, the ε r and σ of skin have been changed from 50% to 150%, and the simulated results are shown in Fig. 15 and Fig. 16, respectively.
When the ε r was increased, the resonant frequency would be shifted to a lower band, as shown in Fig. 15. Meanwhile, it is shown in Fig. 16 that the σ has little impact on the resonant frequency. Besides, the relationship between the    working frequency and ε r could be written as [29] where the f stands for the working frequency, c refers to the speed of light, and l p represents the physical length of MPA. From Eq. (4), it can be seen that the f is inversely proportional to ε r when l p and c are constants, which are well consistent with the phenomenon in Fig 15. According to Eq. (4), the working frequency of MPA could not be directly affected by the σ , as shown in Fig. 16. Moreover, there were obvious variations in impedance matching when the ε r and σ of skin were changed from 50% to 150%. In fact, both of the ε r and σ have a huge influence on the tanδ of skin. Thus, the relationship among these three parameters can be deduced as tan δ = σ ωε 0 ε r (5) where ω = 2πf is the radian frequency. In Fig. 4 and Fig. 5, the MPA impedance is reduced with the increased implant depth, which can decide the drastic decline of Q T in human tissue, shown in Eq. (2). In [35], it was proved that the most important factor of the steep fall of the Q T was brought about by the tanδ of human tissue. That was the reason why the variations of ε r and σ can effectively affect the impedance matching with the tanδ.
Moreover, the embedded location might be changed with human tissue activities and posture movement even if an implantable antenna was designed for a specific IMD. Fig. 17 illustrates the simulated |S dd11 | with different h s . It can be seen that there are almost the same resonant responses with different h s , which are consistent with the phenomenon in Fig. 6 and Fig. 8.
Even though the variations of the dielectric properties (50%) and implant depth (10 mm) of the tissue were larger than those reported in the past, it was found out that the impedance bandwidth of the proposed antenna cloud cover the desired 2.4 GHz ISM band. It is suggested that the proposed MPA had better tolerance to the diversity of human tissue than other reported ones. In addition, the maximized impedance bandwidth can lead to the impressive stabilities of the proposed MPA to various properties of tissue, as discussed above.

D. MODEL INTEGRITY STUDY
As it was mentioned above, the quantitative analysis was realized by the simplified human tissue model and the pacemaker model. Besides, the human tissue was simplified as the one-layered skin model. In addition, the thickness of the pacemaker case or virtual ground plane was set as 0 mm. Practically, the actual IMDs should be used to design the implantable antenna. To realize the performance of the designed MPA assembled with the realistic pacemaker model in the human body, more and more complicated models should be shown in Fig. 18(a) and Fig. 18(b), respectively.
The pacemaker in Fig. 18(a) was modeled as a hollow closed metal (titanium) cavity filled with electronic components, i.e., battery and circuitry. The size of the pacemaker was 40 mm × 40 mm, the thickness of the whole case was set as 10 mm, and the proposed MPA was still printed on the Rogers RO6010 with a thickness of 0.635 mm. Previously, the chest of the Gustav or anatomical human body model places this complicated pacemaker model, as shown in Fig. 18(b), which is also labeled as 'complete model' in Fig. 19. What is more, the proposed MPA in the past with infinitely thin ground embedded in the one-layered skin model is marked as the 'simplified model'. From Fig. 19, it    is seen that the complete model and its two resonant modes still have good resonant performance in the desired operating frequency band.

E. BIOCOMPATIBLE MATERIALS STUDIES
Biologically, the implantable antennas should not only prevent rejection of the implant but also preserve patient safety [1]. Previously, the proposed MPA was directly exposed to the tissue. It was a useful method to insulate the   implantable MPA with a layer of biocompatible coating [1], which made it easy to divide human tissue from direct contacting to implantable devices. Accordingly, zirconium with thin was chosen as the biocompatible superstrate. Fig. 20 shows the schematic geometry of the proposed MPA coating with zirconium. The thickness of the zirconium was marked as h z . The simulated |S dd11 | of the proposed MPA with different thickness of zirconium is shown in Fig. 21. The h z = 0 mm is an indication that the MPA works in the skin with no superstrate. Fig. 21 shows that there is a wide impedance bandwidth in the proposed MPA with a superstrate. Fig. 23 show the photographs of the fabricated MPA and the experimental facilities, respectively. From Fig. 22, it is seen that the proposed MPA is fabricated on the basis of the simplified model. The proposed MPA was characterized through implanting it in the depth of h s = 20 mm inside minced-pork [14], [20], and [22]. Its performance was experimentally confirmed, as shown in Fig. 23.

Fig. 22 and
The pork with an overall dimension of 100 mm × 100 mm × 80 mm was housed by a plastic container. The dual-ports of the differential-driven MPA with equal amplitude and 180 • phase difference [30], [38] was fed with a wideband power divider. The vector network analyzer (VNA) and SATIMO near-field antenna measurement system were used to get the measured |S dd11 | and realize gain patterns of the proposed antenna. In addition, the measurement environment was kept at the ambient temperature of 25 • C, among which 60% was relatively humid. Fig. 24 shows the measured and simulated results of |S dd11 |, accompanied by the photographs of the fabricated prototype of the proposed MPA and the VNA. Nevertheless, it is easy to observe the two resonant modes. Fig. 25 shows the normalized realized gain patterns of simulation and measurement at the first resonant frequency with the photograph of the fabricated MPA in SATIMO. Good agreement between the measured and simulated results is achieved. Besides, some uncertain or unexpected factors can lead to a small amount of variations in the measured results, including measurement tolerance and manufacturing irregularities.

V. CONCLUSION
By fully utilizing the properties of the human tissue in deep implantation to overcome the detuning effect, a design approach is used for the first time to design the implantable microstrip patch antennas in this paper. The influence of the surrounding tissue on the deeply implanted MPA is quantified. Specifically, the Q T of deeply implanted MPA is calculated from the input impedance. Two important parameters, i.e., input impedance and Q T , of MPA, are found to remain unchanged with an extremely low value in deep implantation. Firstly, a robust resonant performance of implantable MPA is realized owing to its stabilities in deep implantation. Secondly, an implantable differential-driven MPA based on dualradiative modes is designed, fabricated, and tested through fully applying the inherent high-loss feature of human tissue, which is helpful to break the limit of the stable and low Q T coming from the influence of human tissue. What is VOLUME 8, 2020 more, the proposed MPA has achieved much wide impedance bandwidth. Next, the sensitivity analysis of the embedded environment is an indication that MPA has an impressive tolerance to various properties of human tissue. The correctness of the proposed theory has been proved based on the model integrity alternatively. In addition, it is found that the simulated results are in agreement with the measured ones.