Anatomical Region-Specific In Vivo Wireless Communication Channel Characterization

TL;DR

This study characterizes in vivo wireless communication channels in the male torso at 915 MHz and 2.4 GHz, revealing region-specific path loss behaviors and multipath effects crucial for healthcare device design.

Contribution

It provides the first comprehensive anatomical region-specific in vivo wireless channel model based on numerical and experimental data at two frequencies.

Findings

Mean path loss decays linearly inside the body.

Power decay rate is approximately twice at 2.4 GHz compared to 915 MHz.

Shadowing variance increases with deeper antenna placement.

Abstract

In vivo wireless body area networks (WBANs) and their associated technologies are shaping the future of healthcare by providing continuous health monitoring and noninvasive surgical capabilities, in addition to remote diagnostic and treatment of diseases. To fully exploit the potential of such devices, it is necessary to characterize the communication channel which will help to build reliable and high-performance communication systems. This paper presents an in vivo wireless communication channel characterization for male torso both numerically and experimentally (on a human cadaver) considering various organs at 915 MHz and 2.4 GHz. A statistical path loss (PL) model is introduced, and the anatomical region-specific parameters are provided. It is found that the mean PL in dB scale exhibits a linear decaying characteristic rather than an exponential decaying profile inside the body, and the power decay rate is approximately twice at 2.4 GHz as compared to 915 MHz. Moreover, the variance of shadowing increases significantly as the in vivo antenna is placed deeper inside the body since the main scatterers are present in the vicinity of the antenna. Multipath propagation characteristics are also investigated to facilitate proper waveform designs in the future wireless healthcare systems, and a root-mean-square (RMS) delay spread of 2.76 ns is observed at 5 cm depth. Results show that the in vivo channel exhibit different characteristics than the classical communication channels, and location dependency is very critical for accurate, reliable, and energy-efficient link budget calculations.

Equations

$$P_r$$

$$PL\left(d\right)=P{L}_0+m\left(d/d_0\right)+S\left(d_o\le d\right)$$

$$\overline{\tau }=\frac{\sum _i{\tau }_{i}P({\tau }_i)}{\sum _{i}P({\tau }_i)}$$

$${\sigma }_{\tau }=\sqrt{\overline{{\tau }^2}-(\overline{\tau }{)}^2}=\sqrt{\frac{\sum _i{\tau }_i^{2}P({\tau }_i)}{\sum _{i}P({\tau }_i)}-{\left(\frac{\sum _i{\tau }_{i}P({\tau }_i)}{\sum _{i}P({\tau }_i)}\right)}^2}$$

$$B_c\approx \frac{1}{50{\sigma }_{\tau }}$$

Full text

I Introduction

Chronic diseases and conditions such as diabetes, obesity, heart disease, and stroke are the leading causes of death and disabilities in the United States. Treating people with these illnesses accounts for 86%111http://www.cdc.gov/chronicdisease of the national health expenditure which is expected to be almost double in the next ten years222https://www.cms.gov. However, these are the most preventable and manageable problems among all health issues by committing to a healthier lifestyle. Continuous health monitoring helps to achieve this goal by assisting people to engage in their healthcare and also allows physicians to perform more reliable analysis by providing the data collected over a large period of time. In addition, exploitation of this big data will replace the traditional “one-size-fits-all” model with more personalized healthcare in the near future. Furthermore, noninvasive surgery and remote treatment are expected to lower the risk of infection, reduce hospitalization time and accelerate recovery processes. All these demanding requirements for an effective service quality in healthcare awakened a general interest in wireless body area networks (WBANs) research [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]. One component of such advanced technologies is represented by wireless in vivo sensors and actuators, e.g., pacemakers, internal drug delivery devices, nerve stimulators, and wireless capsules as shown in Fig.1. In vivo medical devices offer a cost efficient and scalable solution along with the integration of wearable devices and help to achieve the vision of advanced pervasive healthcare, anytime and anywhere [1]. Besides healthcare, the use of in vivo-WBANs is also envisioned for many other applications such as military, athletic training, physical education, entertainment, safeguarding, and consumer electronics [11, 12].

In vivo-WBANs and their associated technologies will shape the future of healthcare considering all the potentials and the critical role of these applications. To fully exploit the use of them for practical applications, it is necessary to obtain accurate channel models that are mandatory to build reliable, efficient, and high-performance communication systems. These models are required not only to optimize the quality of service metrics such as high data rate, low bit-error rate, and latency but also to ensure the safety of biological tissues by careful link budget evaluations. Although, on-body wireless communication channel characteristics have been thoroughly investigated [13, 3], the studies on in vivo wireless communication channels (implant-to-implant and implant-to-external device links) are limited. The in vivo channel exhibits different characteristics than those of the more familiar wireless cellular and Wi-Fi environments since the electromagnetic (EM) wave propagates through a very lossy environment inside the body, and dominant scatterers are present in the near-field region of the antenna.

The IEEE 802.15.6 standard [14] was released in 2012 to regulate short-range wireless communications inside or in the vicinity of the human body. According to this standard, in vivo-WBAN devices operate in the medical device radio communications service (MedRadio) which uses discrete bands within the 401–457 MHz spectrum including the previous specification called medical implant communication service (MICS) band. Despite the fact that MedRadio bands provide satisfying propagation characteristics inside the human body [15], they suffer from lower bandwidths and larger antenna size issues compared to the antennas designed to operate at higher frequencies. Therefore, other frequency bands, such as industrial, scientific, medical (ISM) and ultra-wide band (UWB) communications bands should also be considered in the upcoming standards for in vivo wireless communications. It is also known that EM wave propagation inside the human body is strongly related to the location of the antenna [15, 8] and hence, the in vivo channel should be investigated for a specific anatomical part. For example, the gastrointestinal tract has been studied for wireless capsule endoscopy applications [16], while the heart area has been investigated for implantable cardioverter defibrillators and pacemakers [17]. Although many in vivo path loss (PL) formulas were reported in the literature [18, 17, 19, 4, 5, 7, 6], they do not provide location specific PL model parameters to carry out accurate link budget calculations. Moreover, detailed human body models are crucial in order to investigate the in vivo wireless communication channel. Various phantoms have been designed to simulate the dielectric properties of the tissues for numerical and experimental investigation. The validation of numerical studies with real experimental measurements is required, however performing experiments on a living human is strictly regulated. Therefore, physical phantoms [18, 8, 4] or anesthetized animals [6, 5] are often used for experimental investigations.

This paper presents a numerical and experimental characterization of the in vivo wireless communication channel for male torso considering various anatomical regions. The location dependent characteristics of the in vivo channel are investigated by performing extensive simulations at 915 MHz and 2.4 GHz using HFSS®. A statistical PL formula is introduced, and anatomical region-specific parameters are provided. The multipath propagation characteristics of the in vivo channel are examined by investigating the polarization and analyzing the delay spread, which is of particular importance for broadband applications. In addition to the thorough simulation studies, experiments are conducted on a human cadaver, and the results are compared with the numerical studies. The preliminary results were presented in [20] and [21]. To the best of authors’ knowledge, this is the first study that investigates the in vivo wireless channel for various anatomical regions both numerically and experimentally on a human cadaver.

The rest of the paper is organized as follows. Section II describes the simulation/experimental setup and explains the measurement methodology in detail. Section III presents the in vivo channel characterization based on the numerical and experimental investigation. A statistical PL formula is provided along with the anatomical region-specific parameters, and multipath propagation characteristics are examined thoroughly. Finally, Section IV summarizes the contributions and concludes the paper.

II Simulation and Measurement Settings

II-A Simulation Setup

Analytical methods are viewed as infeasible and require extreme simplifications [2, 22]. Therefore, numerical methods, which provide less complex and appropriate approximations to Maxwell’s equations, are preferred for characterizing the in vivo wireless communication channel. In this study, we used ANSYS HFSS® 15.0 [23], which is a full-wave EM field simulator based on the Finite Element Method (FEM). ANSYS also provides a detailed male human body model, and it includes frequency dependent dielectric properties of over 300 parts (bones, tissues and organs) with 2 mm resolution. This extensive simulation work was beyond the capability of personal computers and advanced computing resources at the University of South Florida (USF) were used to solve such large EM problems. Research Computing at USF hosts a computer cluster which currently consists of approximately 500 nodes with nearly 7200 processors cores and 24TB of memory in total.

The simulations were designed considering an implant to an external device (in-body to on-body) communications scenario in the male torso with a similar measurement setup in [20]. Rather than using the whole body, the torso area was segmented into four sectors considering the major internal organs: heart, stomach, kidneys, and intestine as shown in Fig. 2a. In each region, simulations were performed by rotating receiver (ex vivo) and transmitter (in vivo) antennas together around the body with 22.5° angle increments (Fig. 2b). The ex vivo antenna was placed 5 cm away from the body surface and the in vivo antenna was placed at ten different depths (from 10 mm to 100 mm) inside the body for each ex vivo antenna location. In addition, antennas were placed in the same orientation to avoid polarization losses.

The received power is expressed using the Friss equation (Eq. 1) for free space links[24]:

[TABLE]

where $P_{t}/P_{r}$ represents transmitted/received powers, $G_{t}/G_{r}$ denotes the gain of the transmitter/receiver antenna, $\lambda$ stands for the free space wavelength, $R$ is the distance between transmitter and receiver antennas and $\left|S_{11}\right|$, $\left|S_{22}\right|$ are the reflection coefficients of transmitter/receiver antennas. Unlike free-space communications, in vivo antennas are often considered to be an integral part of the channel [2] (i.e., the gain cannot be separated from the channel) and hence, they need to be designed carefully. However, omnidirectional dipole antennas at 915 MHz and 2.4 GHz were deployed in our simulations for simplicity. The dipole antenna length is proportional to the wavelength, which varies with respect to both frequency and permittivity. Higher frequencies compared to the MedRadio bands provide smaller antenna sizes, hence, they could be implanted conveniently. In addition, the antennas were optimized inside the body with respect to the average torso permittivity for each frequency towards obtaining maximum power delivery. Although the antennas presented a good return loss (i.e., less than -7dB), they were perfectly matched by compensating the $\left(1-\left|S_{11}\right|^{2}\right)$ and $\left(1-\left|S_{22}\right|^{2}\right)$ factors to yield a fair comparison for PL analysis. Also, the mesh size was set to be less than $\lambda/5$ in this study.

II-B Experimental Setup

The numerical investigation was validated by conducting experiments on a human cadaver in a laboratory environment. Istanbul Medipol University provided the ethical approval and medical assistance in this study. The preliminary results were presented in [21]. Animal organs are used to represent human tissues as suggested in [25, 26, 18] and the decayed human internal organs in this experiment were replaced with internal organs of a sheep. The male torso area was investigated at 915 MHz by measuring the channel frequency response, i.e., $S_{21}(f)$, through a vector network analyzer (VNA). A tapered slot coplanar waveguide (CPW)-fed antenna [27] (in vivo) and a dipole antenna (ex vivo) were used in our experiments with two coaxial cables each having a length of 2 meters as illustrated in Fig. 3. The frequency response of cables was subtracted from the channel measurements by performing a calibration of the VNA. The antennas were wrapped with a biocompatible polyethylene protective layer and sealed tightly in order not to contact the biological tissues directly, which could lead to shortening the antennas. The antennas were tested before the experiment and provided sufficient return loss inside the body during the experiments (i.e., less than -7dB).

The in vivo antenna was placed at six different locations (Fig. 4) inside the body around the heart, stomach, and intestine by the help of a physician. In vivo depth measurements were performed using a digital caliper and the antennas were placed with the same orientation to avoid polarization losses, similar to the simulations. The channel data was captured between the frequencies 905 MHz to 925 MHz and post processed for further analysis in MATLAB®. Although the experimental setup did not allow capturing the effects of circulatory and respiratory systems, it provides a more realistic multipath propagation scenario than computer simulations or experiments which are conducted on physical phantoms and anesthetized animals by providing EM wave propagation in an actual human body.

III IN VIVO Channel Characterization

III-A *Path Loss and Shadowing *

The in vivo path loss (PL) expresses a measure of the average signal power attenuation inside the body and is calculated as $PL=-mean\{|S_{21}|\}$ using the channel frequency response, i.e., $S_{21}$ [5, 4]. The location dependent characteristic of the in vivo PL was investigated for two ISM bands, i.e., 915 MHz and 2.4 GHz. The EM wave propagates through various organs and tissues regarding different antenna locations, and the PL changes significantly even for equal in vivo depths. The location dependent characteristic of the channel is more dominant when the in vivo antenna is placed deeper inside the body. Fig. 5 presents the mean PL for the investigated four anatomical body regions in the simulation environment. Although the signal power attenuation is similar for near-surface locations, complex body areas such as intestine cause higher PL due to their dense structure beyond 30 mm in vivo depth.

Various analytical and statistical PL formulas have been proposed for the in vivo channel [1]. Despite the fact that analytical expressions provide intuition about each component of the propagation models, they are not practical for link budget design. According to the final report of the IEEE 802.15.6 standard’s channel modeling subgroup, Friis transmission equation (Eq. 1) can be used for in vivo scenarios by adding a random variation term [28, 29]. In this work, the in vivo PL is modeled statistically as a function of depth by the following equation expressed in dB scale:

[TABLE]

where $d$ represents the depth from the body surface in mm, $d_{0}$ stands for the reference depth with a value of 10 mm, $PL_{0}$ denotes the intercept term in dB, $m$ is the decay rate of the received power and $S$ denotes the random shadowing in dB, which presents a normal distribution for a fixed distance. The power decay rate $(m)$ heavily depends on the environment and is obtained by performing extensive simulations and measurements. Also, the shadowing term $(S)$ depends on the different body materials (e.g., bone, muscle, fat, etc.) and the antenna gain in different directions [17]. The proposed in vivo PL model is valid for $10\leq d\leq 100\,mm$ and the communication channel between an in vivo medical device, and a far external node could be considered as a combination of two concatenated channels: “in-body to on-body” and “classical indoor channel”, if there are no surrounding objects around the body [28]. It should be pointed out that the model is antenna dependent as the majority of other WBAN propagation models in the literature, and this phenomenon is needed to be taken into account for link budget calculations as well.

Figs. 7 shows the scatter plots of PL versus in vivo depth on torso in the simulation environment at 915 MHz and 2.4 GHz. The mean PL is obtained using a linear regression model. It is observed that the power decay rate $(m)$ is approximately twice at 2.4 GHz due to the high absorption in tissues as compared to 915 MHz (Fig. 6). In addition, the variance of the shadowing term, $\sigma$, becomes notably larger as the in vivo antenna is placed deeper inside the body as shown in Table I. This behavior can be interpreted using the fundamental statistics theorem which states that the variance of independent random variables’ sum equals to the sum of the variances of the random variables (scattering objects) involved in the sum. The in vivo channel exhibits a different characteristic than the classical channels, due to the main scatterers present in the vicinity of the antenna, and the variance of shadowing increases significantly compared to free space communications.

The statistical in vivo PL model parameters in Eq. 2 are provided for each anatomical regions in Tables II and III, which were obtained by performing extensive simulations. By interpreting them, it could be concluded that PL increases when the in vivo antenna is placed in an anatomically complex region. For example, the intestine has a complex structure with repetitive, curvy-shaped, dissimilar tissue layers, while the stomach exhibits a smoother structure. As a result, the PL is greater in the intestine than in the stomach even at equal in vivo antenna depths. Also, more radiation occurs in the posterior region than in the anterior region due to the human body structure. To sum up, the location dependency is very critical for link budget calculations and the target anatomical region should be taken into account to design a high-performance, energy-efficient communications system inside the body.

The numerical studies were validated with experiments on a human cadaver at 915 MHz. The in vivo antennas were placed at six different locations as shown in Fig. 4 and the ex vivo antenna was placed 2 cm away from the body surface. Table IV presents the PL values for the selected in vivo locations and comparison of experimental results with numerical studies are provided in Fig. 8. The discrepancies between the simulated and measured results exist due to the additional losses (e.g., antenna distortion), which were not considered in the simulations and the differences in experimental environment. The statistical in vivo PL model parameters are also provided for the experimental study and compared with the numerical study in Table V.

III-B *Multipath Characteristics *

In addition to the PL and shadowing, multipath propagation characteristics of the in vivo channel are also important and should be investigated to discuss proper waveform designs. Received signal strength was explored for various antenna polarizations towards understanding the existence of multipath reflections in the human body medium. As similar to the previous part, the dipole antennas at 915 MHz were deployed in the simulation environment, and they were perfectly matched as mentioned in Section II. The in vivo antenna was placed at 5 cm depth on the chest, and the ex vivo antenna was placed 5 cm away from the body surface to investigate the “in-body to off-body” link. As a baseline to compare with the in vivo channel, the antennas were separated from each other by 10 cm in free space. The ex vivo antenna was rotated with 11.25o increments in the YZ-plane for both scenarios as shown in Fig. 9 and the maximum available power at the receiver for different polarization mismatch angles is presented in Fig. 10 . In the free space link, the received power degrades dramatically as the polarization mismatch increases due to the absence of multipath components, i.e., only line-of-sight components are effective on the received signal strength. On the other hand, the received signal power does not change significantly with polarization mismatch for in vivo medium. Therefore, it can be concluded that biological tissues inside the human body do not absorb the EM waves completely at 915 MHz and allow reflections that lead to multipath propagation. These reflections will cause small-scale fading which is defined as variations over short distances due to constructive and destructive additions of the signals.

As a result of multipath propagation inside the human body, the amount of delay spread should be understood to design an efficient in vivo communications systems. Therefore, power delay profiles (PDPs) for various anatomical regions were extracted from the simulation results. The in vivo antennas were placed at 5 cm depth on the torso, and the ex vivo antennas were placed 5 cm away from the body surface as shown in Fig. 11 for four different directions at 915 MHz. The channel impulse response, $h(t)$, was obtained by taking the inverse discrete Fourier transform (IDFT) of the channel frequency response, $S_{21}$. The PDP was calculated as $PDP(t)=|h(t)|^{2}$ and the total power is normalized to 1 as presented in Fig. 12. Related multipath channel statistics,* mean excess delay *($\overline{\tau}$), and RMS delay spread ($\sigma_{\tau}$) are calculated to quantify the time dispersion effect of the in vivo channel as follows[30]:

[TABLE]

[TABLE]

where $P(\cdot)$ represents the received power in linear scale and, $\tau_{i}$ denotes the arrival time of the $i^{th}$ path. These parameters for various anatomical directions are listed in Table VI and it is observed that the maximum difference in $\sigma_{\tau}$ is 0.3 ns. Therefore, it can be stated that there is almost no difference in delay spread for various locations when the antennas are implanted with 5 cm depth on the torso.

RMS delay spread determines the coherence bandwidth ($B_{c}$) of the channel. It is a statistical measure of the range of frequencies where the channel can be assumed as “flat” [24] and the 90% $B_{c}$ is estimated as follows:

[TABLE]

The average $\sigma_{\tau}$ at 5 cm in vivo depth is measured as 2.76 ns on the torso and 7.25 MHz coherence bandwidth was predicted using Eq. 5. Theoretically, inter-symbol interference (ISI) is not a critical problem when the signal bandwidth (BW) is less than $B_{c}$. Hence, the measured delay spread should not cause serious ISI for narrow-band (NB) communications. However, this dispersion may lead to a significant interference for UWB communications, which occupies a BW of greater than 500 MHz.

In frequency-selective channels (i.e., the signal BW is greater than $B_{c}$) single-carrier waveforms might not exhibit a sufficient bit error rate (BER) performance without deploying complex equalizers to solve the ISI problem. Nevertheless, power limitation is a major constraint for in vivo-WBAN devices and hence, the complexity of signal processing operations must be low. Multi-carrier systems are offered to provide a trivial solution for the ISI problem. For example, orthogonal frequency division multiplexing (OFDM) based waveforms can easily handle delay spread using a cyclic prefix. However, high peak-to-average-power ratio (PAPR) emerges as a common problem in multi-carrier waveforms, and it makes the signal vulnerable against the non-linear characteristics of the radio frequency (RF) front-end components. Since the in vivo-WBAN devices are restricted in size, the use of high-quality components with high dynamic ranges is impractical. Therefore, PAPR remains as an important issue and may still lead the designers to use single-carrier signaling techniques. To sum up, there are tradeoffs in waveform selection considering the dispersive nature of the in vivo channel and practical issues together. The system requirements in terms of throughput, power efficiency, and signal quality need to be clearly identified, and the most proper waveform technology should be selected accordingly.

IV Conclusions

This paper presented the location dependent characteristics of the in vivo wireless communications channel for male torso at 915 MHz and 2.4 GHz. Extensive simulations were performed using a detailed 3D human body model and measurements were conducted on a human cadaver. A statistical in vivo path loss model is introduced along with the anatomical region-specific parameters. It is observed that the path loss in dB scale follows a linear decaying profile instead of an exponential characteristic, and the power decay rate is approximately twice at 2.4 GHz as compared to 915 MHz. In addition, the variance of shadowing increases significantly as the in vivo antenna is placed deeper inside the body since the main scatterers are present in the vicinity of the antenna. Results show that the location dependency is very critical for link budget calculations, and the target anatomical region should be taken into account to design a high-performance in vivo communications system without harming the biological tissues. Multipath propagation characteristics are examined as well to facilitate proper waveforms inside the body by investigating various antenna polarizations and PDPs. A mean RMS delay spread of 2.76 ns is observed at 5 cm in vivo depth. Despite the fact that this dispersion may not cause significant ISI for NB communications, it could be a serious issue for UWB communications. The interest in WBANs is rapidly growing and in vivo medical devices are shaping the future of healthcare. This study will contribute significantly to the upcoming WBAN standards and hence, will lead to the design of better in vivo transmitter/receiver algorithms.

ACKNOWLEDGMENT

This publication was made possible by NPRP grant # NPRP 6 - 415 - 3 - 111 from the Qatar National Research Fund (a member of Qatar Foundation). The authors are thankful to Istanbul Medipol University, School of Medicine for providing the human cadaver and their valuable medical assistance.