Fat-IntraBody Communication at 5.8 GHz: Verification of Dynamic Body Movement Effects Using Computer Simulation and Experiments

This paper presents numerical modeling and experimental validation of the signal path loss at the 5.8 GHz Industrial, Scientific, and Medical (ISM) band, performed in the context of fat-intrabody communication (fat-IBC), a novel intrabody communication platform using the body-omnipresent fat tissue as the key wave-guiding medium. Such work extends our previous works at 2.0 and 2.4 GHz in the characterization of its performance in other useful frequency range. In addition, this paper also includes studies of both static and dynamic human body movements. In order to provide with a more comprehensive characterization of the communication performance at this frequency, this work focuses on investigating the path loss at different configurations of fat tissue thickness, antenna polarizations, and locations in the fat channel. We bring more realism to the experimental validation by using excised tissues from porcine cadaver as both their fat and muscle tissues have electromagnetic characteristics similar to those of human with respect to current state-of-art artificial phantom models. Moreover, for favorable signal excitation and reception in the fat-IBC model, we used topology optimized waveguide probes. These probes provide an almost flat response in the frequency range from 3.2 to 7.1 GHz which is higher than previous probes and improve the evaluation of the performance of the fat-IBC model. We also discuss various aspects of real-world scenarios by examining different models, particularly homogeneous multilayered skin, fat, and muscle tissue. To study the effect of dynamic body movements, we examine the impact of misalignment, both in space and in wave polarization, between implanted nodes. We show in particular that the use of fat-IBC techniques can be extended up in frequency to a broadband channel at 5.8 GHz.

2.4 GHz or 5.8 GHz are typically chosen for in-body implants as a compromise between small antenna size (higher frequencies allow smaller antennas) and acceptable penetration through human body tissues, although at lower frequencies the penetration is higher [3]. One main challenge for in-body implants is the miniaturization of the antenna. Because of the lossy nature of biological tissues, higher-frequency waves attenuate faster than lower-frequency waves, and therefore reduce the communication range. Numerous implantable systems are currently available in the market, including drug delivery systems [4]- [6], cochlear implants [7], [8], pressure monitors in the brain [9]- [11], glaucoma sensors [12], [13], and direct control of prostheses [14]- [16]. By including the wireless technology in their market offers, the use of such systems is enormously extended. This trend is also widely noticed in the literature through the increasing interest in miniature antennas for implants [17]- [21].
In previous studies at lower microwave frequencies (2.0 and 2.4 GHz) [22]- [24], the path loss of fat-intrabody communication (fat-IBC) has been investigated against different factors that might affect the signal path loss. These studies show that the path loss depends on several properties of the fat tissue. When there are perturbants or muscles that correspond to a height greater than 40% of the fat's thickness, the path loss increases significantly [25], while for the physical curvature variation of fat tissue, a loss of 4 dB is expected in extreme scenarios [26]. Blood vessels in fat tissue cause a path loss that depends on their position and orientation in the fat tissue; they can behave as a reflector, a splitter, or as a splitter and re-combiner [27]. Recent studies report on the effect of steel wires, placed in human chest in the context of open-heart surgeries, on microwave intrabody communication [28]. Särestöniemi et al. discuss intrabody channel characteristics between a capsule endoscope and a receiving directive on-body antenna at UWB frequencies for wireless body-area networks [29]. Inspired by the insights gained from these previous studies and the value of considering operations at higher frequencies, in this work we investigate fat-IBC at 5.8 GHz using numerical simulations and laboratory experiments using porcine cadaver tissues to enhance realism.
Moreover, we propose a small probe optimized to operate efficiently into the fat tissue at a frequency of 5.8 GHz. The small size of the probe, together with the experiment setup, makes it possible to investigate new aspects of this intrabody communication. In particular, the small size allows the study of intrabody communication subject to dynamic body movements where the position of implanted nodes are likely to be displaced. The displacement of nodes can change the relative orientation and position between nodes as well as their respective polarization, affecting the quality of the channel. These investigations have the potential to provide valuable information that can be utilized in designing reliable communication between implants, particularly for fat-IBC. Table 1 shows the difference in probe size between some previous work at 2.0 GHz and 2.4 GHz and the current work. The novelty of the research is that the use of fat-IBC techniques can be extended up in frequency to a broadband channel at 5.8 GHz. We study the effect of dynamic body movements by examining the impact of misalignment, both in space and in wave polarization, between implanted nodes. Moreover, another novel aspect of this paper is the development of a simplified generic path loss (PL) model, which describes the signal attenuation between a transmitter (TX) and a receiver (RX) antenna as a function of the propagation distance at 5.8 GHz. We select the 5.8 GHz frequency corresponding to the ISM band, as this choice is one way to reduce the antenna size and to make the antenna feasible for implantation. Another advantage of this choice is the wide bandwidth offered by operating at 5.8 GHz, which enables high data rates that could be utilized, for instance, for live streaming of gastro-intestinal video signals.
The paper is organized as follows. Section 2 describes the dielectric properties of the ex-vivo tissues and a proposed phantom model used in the laboratory and compared to the human tissue. The description of the topology optimized probes, along with the measurement setup of the probes, is given in Section 3. In Section 4, we describe the experimental setup for a set of possible configurations for the fat channel and different alignments between the probes. Section 5 presents numerical and experimental results that demonstrate the benefit of our new investigation of the fat channel at 5.8 GHz. The conclusion is given in Section 6.

II. DIELECTRIC PROPERTIES OF THE TISSUE-EQUIVALENT PHANTOM, EX-VIVO, AND HUMAN TISSUES
In this study, we perform experiments in two environmentstissue-equivalent phantom and ex-vivo tissue-and compare with the human tissue developed by Andreuccetti et al. [30], based on the work of Gabriel et al. [31], [32]. Tissue-equivalent phantoms emulating skin and muscle were made as agar-based compounds, while those emulating fat were made as oil-based compounds. The phantoms used in this work were developed in our laboratory using recipes similarly as done in Joseph et al. [33]. For the ex-vivo tissues, the skin, fat, and muscle were separated from the porcine belly portion obtained from a local slaughterhouse. The fat and muscle have been finely minced with a meat mincer to reduce air traps between the two layers when forming the model structure.
To establish suitable models to study fat-IBC, the frequency-dependent, real ε and imaginary ε , parts of the dielectric permittivity of the skin, fat, and muscle tissues are determined. We measure these quantities using a Keysight N1501A open-ended dielectric slim form probe kit. For numerical simulations, Cole-Cole models are used to fit the measured data. Figure 1 shows a comparison between the dielectric properties, over the frequency range 3-8 GHz, of the three models: (i) freshly excised ex-vivo porcine belly tissues, (ii) tissue-equivalent phantom compounds, and (iii) human tissues. From Figure 1a we can see that the dielectric constant ε of the ex-vivo skin tissue is slightly lower than that of the human skin. However, we could consider them to be statistically similar, as the reported uncertainty in the Keysight N1510A dielectric probe system is about 10% [34], mainly at lower frequencies, and as they are tissues of different nature. Moreover, the uncertainties from the data in the IFAC database, Andreuccetti et al. [30], could be considered in the worst-case scenario as much as 10%, according to the work by Gabriel and Peyman [35]. Figure 1b presents the dielectric properties of the fat tissue as a function of frequency. It shows that the dielectric properties of an ex-vivo fat tissue are similar to the human fat tissue. The dielectric constant, ε of the oil-based phantom is also similar above 5 GHz, but ε is twice the value of ex-vivo and human tissue. This is a drawback of the phantom recipe that was employed to ensure a flexible phantom that could be used in experimental scenarios where the phantom needs to be bended. Further, this phantom is used to study the behaviour of the fat channels possessing different electromagnetic properties. The muscle tissues of the phantom and ex-vivo have smaller ε compared with those of the human tissue, whereas their ε have slightly higher values than those of the human tissue (see Figure 1c). Therefore, these values can be used in our experiments. Table 2 shows the comparison of the dielectric properties extracted at 5.8 GHz ISM band from Figure 1. Since the fat tissue has a lower loss tangent tan δ, and since waves in this medium travel faster than in the muscle and skin tissues, a wave travelling through the fat will experience a lower path loss than in other tissues. It can be seen that the ex-vivo porcine fat tissue has similar characteristics as the human tissue, while the phantom has twice their loss tangent.

III. TOPOLOGY OPTIMIZED COMPACT PROBE A. PROBE OPTIMIZATION
To investigate the fat channel communication at 5.8 GHz, we need to efficiently launch and record electromagnetic waves into and from the fat tissue. To achieve this goal, we design a wideband probe by using our topology optimization approach [36]. Figure 2 shows the configuration of the proposed probe, consisting of a coaxial cable connected to one side of a rectangular cavity with dimensions W probe = 19.05 mm T probe = 9.525 mm and L probe = 14.3 mm. The cavity region is filled with a substance having permittivity properties similar to that of the fat tissue. The opposite side of the cavity is open towards the fat tissue domain ex and represents a radiating aperture with a size of W probe × T probe . The aperture size is similar to the dimensions of a standard WR75 waveguide, which allows the use of WR75 waveguide sections for the probe assembly. Outside the cavity, the domain ex has a wave impedance that depends on the VOLUME 9, 2021 TABLE 2. Comparison between the dielectric properties of the tissue-equivalent phantom, ex-vivo porcine tissue, and human tissue at 5.8 GHz. complex permittivity of the medium. Here, domain ex is assumed to possess the complex permittivity of the fat tissue. Note that the matching between the probe and ex depends on the probe's geometrical parameters, the material property inside the probe, and the wave impedance of the exterior domain, as will be demonstrated next.
To design the probe, we formulate the conceptual optimization problem where σ is the conductivity distribution in the planar domain , located at the middle of the cavity; W out,coax | W PWy is the received energy in the coaxial cable when a y-polarized plane wave propagates from ex towards the probe; and W out,coax | W in,coax is the reflected energy to the coaxial cable when the coaxial cable is the source of transmission. Problem (1) reads: find the conductivity distribution in that (i) maximizes the energy received by the probe W out,coax | W PWy when the probe operates in the receiving mode and, at the same time, (ii) minimizes the energy reflected by the probe W out,coax | W in,coax when the probe operates in the transmission mode.
We solve problem (1) using gradient-based optimization. The finite-difference time-domain (FDTD) method [37] in an in-house developed software numerically solves Maxwell's equations to compute the two factors of the objective functions. We use a uniform FDTD grid with a spatial step size = 0.10583 mm, and a temporal step size t equals 0.9 of the Courant stability limit, and 10 cells of a uniaxial perfectly matched layer (UPML) to truncate the simulation domain in all directions. The Coaxial cable port is simulated using the model presented in our previous work [38]. The domain is pixelated into 90 × 135 Yee cells containing 24 525 edges. The vector of design variables holds the conductivity value at each edge in . The gradient of the objective function is computed by an expression involving FDTD solutions of the governing Maxwell equations together with an adjoint system of equations associated with the current objective function [39]. The frequency band of interest is 3.6-7.2 GHz, which is determined by controlling the frequency spectrum of the excitation signals. The vector of design conductivities is updated iteratively by the globally convergent method of moving asymptotes (GCMMA) [40]. The GCMMA is a gradient-based optimization method particularly suitable for large-scale topology optimization problems. More details about the numerical treatments of problem (1) are presented in previous work [36], [41]. Figure 3 shows the development of the objective function and the design in versus iteration number. The optimization algorithm converged after 156 iterations to the design shown in the inset of Figure 3. The black and white colors indicates copper and dielectric, respectively. We emphasize the large scale of the optimization problem: each pixel in the final design is a design variable, and there are 24 525 pixels. We present the design as an image with its axes scaled in millimeters to facilitate design reproducibility.

B. DESIGN AND FABRICATION OF THE PROBES
A prototype of the optimized probes is designed, developed, and fabricated. The probe consists of four elements: A rectangular waveguide section, a 50 SubMiniature version A (SMA) connector, a topology optimized planar antenna, and a dielectric filling with a similar permittivity as the fat tissue. Figure 4 shows the prototype of the fabricated probe with the topology optimized planar antenna immersed inside the fat-equivalent phantom material. Three-dimensional (3D) perspective view of the probe is shown in Figure 4a. The topology optimized planar antenna (TOPA) is printed on a single-layer printed circuit board with a size of only 14.29 mm × 9.52 mm (see Figure 4b). The copper plate material is soldered in one side to the broad wall of the waveguide section. The TOPA was fabricated using 0.64 mm thick Rogers TMM6 substrate with dielectric constant ε' of 6 ± 0.08, loss tangent tan δ of 0.0023, and 0.035 mm copper thickness (see Figure 4c).
The performance of the optimized probe is analyzed by our in-house finite-difference time-domain (FDTD) code and cross-verified using the commercial CST Microwave Studio software package. We use the transient CST solver which employs the finite integration technique to numerically solve the electromagnetic governing equations [42]. As discussed in Section III-A, the performance of the probe depends on the properties of the surrounding medium. Here, we demonstrate this fact by comparing the performance of the probe when it radiates into different environments. The probe's performance is investigated based on computing the reflection coefficient when the probe is attached to the phantom-equivalent fat tissue (the medium for which the probe is optimized) and when it radiates into free-space (for which the probe is not optimized). Figure 5 shows the simulated reflection coefficient of the probe in both environments for comparison. When the probe is attached to the fat tissue, the |S 11 | is lower than -12 dB over the frequency range of 3.2-7.1 GHz, which contains the entire F-band frequencies of 4.9-7.05 GHz. These results show that the probe indeed has an excellent match with the fat tissue at 5.8 GHz. However, the |S 11 | is essentially above -5 dB when the probe radiates into free space. This increase in |S 11 | indicates that more energy is reflected back to the source. That is, the probe's performance deteriorates if it radiates into a medium different from what it was optimized for. The wideband performance of the optimized probe can be used to investigate the performance of the fat tissue at neighboring frequencies as well. We emphasize  that our probe is used only to characterize the intrabody fat communication. For in-body or on-body communication, antennas with other objectives and constraints are typically needed [43], [44].    Figure 6 shows the probe-to-probe measurement setup. The two probes, acting as a transmitter (TX) and a receiver (RX), are in contact with each other. The TX is connected to Port 1 and the RX to Port 2 of the microwave analyzer. The scattering parameters of the probes were measured using the Keysight Microwave FieldFox Analyzer (N9918A) [45] in the 3-8 GHz frequency range with data acquired at 1001 frequency points. A microwave absorber has been used to reduce the interaction between the probes and the surrounding environment. Figure 7 shows the probe-to-probe simulated and measured scattering parameters. The simulated and measured coupling coefficients |S 21 | of the probe-to-probe setup are -2.0 dB and -2.2 dB, respectively, at 5.8 GHz. The probes show a reflection coefficient |S 11 | below -10 dB over the frequency range 4.4-8.0 GHz and 4.7-8.0 GHz for simulations and measurements, respectively. The small differences between the simulation and measurement results can be ascribed to air gaps in the dielectric material inside the probes as well as air gaps between the two probes. Figure 8 shows the Voltage Standing Wave Ratio (VSWR) of the probes. As we can see, the VSWR is below two at the frequency range of 4.3-8.0 GHz for simulation and 4.8-8.0 GHz for the measurement. These results confirm that the proposed probes are matched to the fat tissue, and that they contribute insertion losses of 2.0 dB and 2.2 dB when the channel path loss is evaluated in simulation and measurement, respectively.

A. EXPERIMENTAL SETUP
A comprehensive setup has been modeled, designed, and developed to characterize the fat-IBC at 5.8 GHz ISM band. In this section, four different configurations, Type A, Type B, Type C, and Type D of the three-layered tissue model structure, are proposed and discussed. All measurements are conducted in two different environments, tissue-equivalent phantom, and ex-vivo porcine tissue. In addition, we use the CST software to simulate these environments. The illustration of each type is shown in Figure 9. The thickness of skin and muscle tissue are fixed in all cases to 2 mm and 30 mm, respectively.
The thicknesses were chosen based on our previous work and they also present the average thickness of these tissue layers [22], [23], [46], [47]. The fat in this context is the tissue that is utilized for the fat-IBC characterization and is varied in two parameters, the thickness and the width. The thickness is changed from T fat = 9.93 mm to T fat_extend = 25 mm, and the width is changed from W = 19.10 mm to W extend = 50 mm. The channel length, L is varied from 20 mm to 100 mm in 20 mm steps.
In Type A, the channel is characterized by having the same L, W , and T fat according to the dimension of the probes (see Figure 9a). In Type B, the width of the channel is extended by W extend = 50 mm, while the other parameters remain the same as in Type A (see Figure 9b). Instead of having extra width, the fat thickness is extended in Type C, while the fat width remains the same as Type A (see Figure 9c). In Type D, we extend both the thickness and width of the fat channel (see Figure 9d). Figure 10 shows the setup for the Type D model when one of the probes is rotated by 90 • in the fat tissue with a 100 mm channel length. The purpose of the latter setup is to observe the signal transmission and analyze the impact of changing the wave polarization on fat-IBC. Overall, the primary purpose of these experiments is to obtain a generic insight on how the path loss characteristics change under different conditions, as well as to understand  the influence of the misalignment, in space or polarization, between the transmitting and receiving probes (nodes).
The experimental setups of the tissue-equivalent phantom and the ex-vivo tissue environment are shown in Figure 11. The phantom and ex-vivo tissues were kept at room temperature in the laboratory at 22 • C (71.6 • F) for at least 30 minutes to equalize their temperature to the room temperature. The experiments are done in room temperature in order to obtain reliable data and to ensure repeatability. The reference cadaver tissue data has been collected at room temperature and our experimental setup complies with the reported values in [48]. Even though the phantom and ex-vivo tissues were stabilized to room temperature, they may still differ in temperature with respect to the human data reported in [48] since the data obtained at temperature as low as 20 • C.
As shown in Figure 11a, the setup consists of a pair of rectangular probes, a Keysight Microwave FieldFox analyzer (N9918A), a coaxial cable connector, and a threelayered skin, fat, and muscle of ex-vivo porcine tissue (see Figure 11b) or phantom (see Figure 11c). The probes and ex-vivo tissue are supported by a 3D printed holder molded from polylactic acid (PLA) plastic material, which has the dielectric properties of 2.6 ± 0.2 in order to provide a stable fixture during the experiment. The frequency range of the measurements is 3-8 GHz. For each reading, we record the scattering parameters based on the average of five measurements.

B. NUMERICAL MODELING SETUP
We study numerically the effect on fat-IBC of misalignment between the probes (nodes) and the change in the wave polarization. We use the CST transient solver to simulate these environments. The simulation models are assigned with the respective dielectric properties of the constituent materials measured during the experiments. A fourth-order polynomial is used to fit the dielectric properties of the measured material data to the simulation model in CST. In our numerical model, we assume a free space medium above the skin layer, and we use a radiation boundary condition to surround the model in all sides to avoid reflections from the boundaries. First, to study the impact on the signal coupling with respect to the alignment, we change the positions of the transmitting and receiving probes vertically as well as horizontally inside the fat channel. We mark the center of the channel as the reference position. Figure 12 shows a model consisting of two probes acting as transmitter (TX) and receiver (RX) that have been implanted VOLUME 9, 2021 in the fat tissue of the three-layered homogeneous tissue structure. The transmitter probe is aligned at four orientations: Up, Down, Left, and Right (see Figure 12a). Figure 12b shows one example where the transmitter is placed on top of the muscle tissue, and the receiver is situated at the center of the fat tissue. Second, we study the impact using a horizontal polarized electric field inside the fat channel. Compared to our previous probes [22], [23], the compact size of the new probes makes it feasible to investigate the use of a horizontally polarized electric field inside the fat channel. To do so, we rotate both the transmitting probe and the receiving probe by 90 degrees inside the fat channel.
The width and the length of the entire tissue are 80 mm × 180 mm, and the distance between the probes is fixed at 100 mm. The thicknesses of the skin, fat, and muscle are 2 mm, 25 mm, and 30 mm, respectively. The aim of the study is to observe the impact on the signal coupling with respect to the misalignment of the probes in the fat tissue, which helps to investigate the influence of misalignment between implant devices.

A. CHANNEL PATH LOSS AT 5.8 GHz
Following the procedure in Section IV-A, we measured and computed the transmission coefficient |S 21 | as indicated in Figures 13-16, where the results of the ex-vivo porcine tissue environment and the tissue-equivalent phantom environment are presented in the left and right graphs,       Figure 13 shows the |S 21 | of Type A simulated and measured results for the ex-vivo and the phantom for different channel lengths. Figure 13a shows an excellent match between the simulated and measured |S 21 | results. Using the change in the |S 21 | values for different channel lengths, we estimate an average path loss of 6 dB/20 mm at 5.8 GHz.
Since the loss tangent of the fat-equivalent phantom is twice that of the ex-vivo tissue, it was estimated that the average path loss is 12 dB/20 mm (see Figure 13b). In our measurements, we use a N9918A FieldFox Handheld Analyzer, which has a maximum dynamic range of -45 dBm across the frequency band of interest. In all figures, we plot the dash-dotted red line at -45 dB to mark the accuracy limit of the analyzer. As can be seen in Figures 13b-16b, the measurement data for 80 mm and 100 mm is almost flat over the entire band of interest. We therefore consider measurement results below -45 dB as faulty measurements and rely on the simulated results to draw our conclusions. Figure 14 shows the |S 21 | simulated and measured results for the ex-vivo and the phantom in the case of additional width in the fat transmission channel (Type B). Similar to Type A, the simulation and  measurement results indicate an average channel path loss of almost 6 dB/20 mm, and the fat-equivalent phantom indicates 12 dB/20 mm. Figure 15 shows the transmission coefficient of the Type C model, which has a thicker fat tissue. As can be seen in Figure 15a, at 100 mm distance, the signal coupling increased by more than 10 dB compared to Figures 13 and 14. Similarly, by comparing Type C results at 60 mm channel length, we note that there is an increase in the values of |S 21 | by 2.5 dB and 4.6 dB compared to Type A and Type B, respectively (see Figure 15b). Figure 16 shows the |S 21 | when the fat tissue is extended around the probes for both environments, ex-vivo porcine tissue and tissue-equivalent phantom. Type D, where the fat tissue surrounds the implant, is the case that mimics real-life scenarios most accurately. With the small probe size, we can easily implant the probe inside the fat tissue. Overall, we observed that Type C and D do not differ much in their signal transmission. This finding emphasizes that fat thickness has a significant influence on signal transmission. Figure 16a shows that the ex-vivo simulated and measured |S 21 | at 100 mm channel length is -25.6 dB and -25 dB, respectively. The simulated and measured |S 21 | for the phantom at 60 mm channel length is -45 dB and -44. dB, respectively (see Figure 16b). Table 3 and 4 show a summary of the simulated and measured transmission coefficient |S 21 | for the different configuration types with respect to the channel length for ex-vivo porcine tissue and tissue-equivalent phantom at 5.8 GHz, respectively. Accordingly, we plot the values of the average path loss (excluding the probe-to-probe insertion loss of 2.2 dB), as shown in Figure 17. Figure 17 shows that the different environments have different path loss profiles.
Overall, our experiments show that Type C and D show lower signal path loss compared to Type A and B. The increase of the fat thickness in Type C and D explains the difference with respect to the other two types. The increase in the fat thickness decreases the path loss. We also observed that all measurement results are in good match with the simulation results. By considering the average for all distances, an estimated path loss per centimeter for the ex-vivo porcine tissue and tissue-equivalent phantom environments are shown as below. The bracketed values are the simulated results, and non-bracketed values are the measured results.
• Type D: -3.8 dB/cm (-3.6 dB/cm) for ex-vivo; -7.5 dB/cm (-7.3 dB/cm) for phantom.  Figure 12a. We note that the entire probe is surrounded by the fat tissue. The TX probe location is varied to study the impact on the signal coupling when nodes are misaligned/displaced in the fat channel medium. As a baseline for our experiments, we use the setup where the two probes are located in the center of the fat tissue. Figure 18 summarizes the impact of the probes' vertical displacement on the signal coupling through the fat channel. Figure 18a shows that the displacement of the transmitting probe towards the skin (Up) or towards the muscle tissue (Down) has only a small impact on the probe's reflection coefficient |S 11 |. Figure 18b shows that the displacement of the transmitting probe down towards the muscle leads to more loss compared to the displacement up towards the skin.   Excluding the results of the 20 mm channel, we conclude that the vertical displacement of the probe increases the channel insertion loss by about 3 dB and 1 dB when the probe (node) is displaced down towards the muscle and up towards the skin tissues, respectively. The increase of the channel insertion loss as the probe is displaced down is attributed to the high losses caused by the muscle tissues near the probe channel interface. Figure 19 summarizes the impact of the probe's horizontal displacement on the signal coupling through the fat channel. Figure 19a shows that the horizontal displacement has essentially no impact on the probe's reflection coefficient (excluding the minor changes for the 20 mm channel length). Moreover, the results of the transmission coefficient show that even at 15 mm horizontal displacement the insertion loss is increased by at most 2 dB compared to the baseline case. Note that the fat channel is horizontally symmetric, which is why the results of the right and left displacements in Figure 19 coincide.   respect to the reference (Type D model) for the simulation and measurement, respectively, at 100 mm channel length.

B. IMPACT OF NODE DISPLACEMENT
To study the impact of imposing a horizontally polarized electric field into the fat channel, the two probes are VOLUME 9, 2021 rotated. The analogy between the fat channel and parallel plate transmission line suggests that this horizontal electric field polarization is likely to excite the TE modes inside the fat channel [49]. Figure 21 shows a comparison between the performance of the fat channel for the two cases.
Similar to the case when the probe imposes a vertical polarized electric field, Figure 21a shows that the probe has a good matching to the fat channel. The |S 11 | is below -15 dB at 5.8 GHz. Figure 21b shows the results of |S 21 | when the distance between the TX and the RX varies from 20 mm to 100 mm. We note that the |S 21 | at 5.8 GHz do not vary much when the probes are rotated 90 degrees. In addition, we notice that the |S 21 | curve is essentially flat as the frequency increases. However towards higher frequencies, the case when both probes are rotated by 90 • (horizontal electric field), the |S 21 | shows 5 dB better coupling compared to the normal position (vertical electric field). This observation suggests that fat-IBC can be further extended towards higher frequencies.
To investigate the near-field performance of the probe, we plot the electric field distribution when the probes are used to impose a vertical and a horizontal electric field into the fat channel, as shown in Figures 22 and 23, respectively. The figures demonstrate the performance of the probe around the 5.8 GHz and two additional plots for the 4.8 GHz and 6.8 GHz which reside at the edge of the F-band frequencies (4.9 GHz-7.05 GHz). Comparing the plots in both figures, we observe a higher signal amplitude for the 90 • rotated case, both in the side-view and top-view for all frequencies, close to the receiver (RX) side. In addition, the comparison of the side-view in both figures shows that there is less leakage close to the transmitting (TX) probe for the case when the probes are 90 • rotated.

VI. CONCLUSION
In this paper, we have investigated the fat-intrabody propagation channel at 5.8 GHz in the ISM band. The signal propagation setups were numerically modeled and developed for six different types, including the effects of probe (node) displacement and a 90 • rotation of the transmitter. The laboratory experiments were conducted in ex-vivo porcine and tissue-equivalent phantom environments. The scattering parameters were determined, and the path loss was estimated from the conducted measurements. The measurement results can be summarized as follows. Firstly, the path loss is dependent on the surrounding fat environment. In ex-vivo model, the increase in the fat thickness causes lower path loss in the channel transmission. Secondly, the change in the wave polarization between the transmitting and receiving nodes can have a severe negative impact on the signal coupling. In the extreme case when one of the node misaligned by 90 • , this can reduce the received signal by around -38.4 dB (measurement) and -35 dB (simulation). Thirdly, the misalignment between two nodes inside the fat channel does not strongly affect the signal coupling, which is as low as 1 dB and 2 dB for both cases. The flat response of the transmission coefficient over a wide bandwidth suggests further investigations of the fat-IBC at even higher frequencies. Future research will include analytical channel modeling, studying the data-rate relation to the low and high frequencies, investigating realistic implant antennas, and more measurements for path loss analysis.

ACKNOWLEDGMENT
The computations were partially performed on resources provided by the Swedish National Infrastructure for Computing (SNIC) at the HPC2N Center. MARTIN BERGGREN received the Ph.D. degree in computational and applied mathematics from Rice University, Houston, TX, USA, in 1996. He is currently a Professor of scientific computing with Umeå University, Umeå, Sweden. Previously, he held a lectureship position at Uppsala University, Uppsala, Sweden, and research positions at the Swedish Defence Research Agency, Stockholm, Sweden, and Sandia National Laboratories, Albuquerque, NM, USA. His research interests include numerical methods for partial differential equations, and, in particular, computational design optimization for acoustic and electromagnetic wave propagation problems.
THIEMO VOIGT (Member, IEEE) received the Ph.D. degree from Uppsala University, Sweden, in 2002. He is currently a Professor of computer science with the Department of Information Technology, Uppsala University. He also leads the Networked Embedded Systems Group, RISE Computer Science. His current research interests include system software for embedded networked devices and the Internet of Things. His work has been cited more than 14 500 times. He is a member of the editorial board for the IEEE INTERNET OF THINGS and ACM Transactions on Sensor Networks (TOSN).